The gig economy presents Marxist class analysis with its most awkward contemporary case, and the awkwardness is not a debating trick. A driver who carries passengers through a ride hailing application owns the car. A courier who delivers meals owns the bicycle, the phone, and often the insulated bag. A cleaner booked through a platform owns the mop, the bucket, and the chemicals in the cupboard under the stairs. In the classical picture, ownership of means of production is what separates the petty bourgeois producer from the proletarian, who owns nothing but labour power and must sell it to live. The platform worker appears to own something real, and that appearance is the first objection every thoughtful reader raises against calling such workers working class. If the objection cannot be answered, the analysis fails at the first step, and everything built on it, claims about exploitation, about algorithmic control, about collective organisation, collapses into assertion. This article answers the objection by doing the analytical work in the open. It asks which means of production actually matter for a given labour process, distinguishes three serious answers to the ownership puzzle, and tests each answer against five kinds of platform work. The conclusion it defends is that class position is settled by ownership of the means of production that are decisive for the labour process, not by ownership of any means of production whatever, and that applying this decisive-means test places platform workers in the working class while explaining precisely why the question felt hard.

The exercise is worth doing carefully because platform work is where the vocabulary of class is used most loosely in public argument. Advocacy writing calls every courier a worker and every platform a boss, as though the words settled the matter. Industry writing calls every courier an entrepreneur and every platform a marketplace, as though those words settled it the other way. Labour law writing, which dominates the accessible coverage, treats the question as a dispute about statutory definitions, and it is, but the statutory question and the sociological question are different questions, and an answer to one is not an answer to the other. A court can decide that a driver is an employee for the purposes of a minimum wage statute while the analytical question, where this person stands in the relations of production, remains untouched. Conversely, a legislature can declare drivers independent contractors by statute without changing by one iota who owns the dispatch system, who sets the price, or who can be excluded from the market. This article keeps the two questions separate throughout, and it treats the legal material as evidence about how states have responded to a relation, not as a verdict on what the relation is.

Three clarifications belong at the front. First, the article works with the Marxist criterion of class in its strict form, set out in the series treatment of the marxist theory of class guide: class position is a relation to the means of production, not an income band, a lifestyle, a level of education, or a self description. Second, the article assumes the account of exploitation developed elsewhere in the series and concentrates on the classificatory puzzle, which is the part of the problem that is genuinely new. Third, the article is honest about scale from the outset. Platform work is a small share of total employment in most economies, far smaller than its share of public argument, and the analysis is offered because the case is theoretically revealing, not because the sector is numerically dominant. A framework that can classify the hard case correctly has earned something. A framework that needs the hard case to be enormous before it matters has earned nothing.

The phenomenon described on its own terms

Before any category is applied, the phenomenon deserves description in neutral terms, because much of the dispute is smuggled in through descriptions. Platform work, as the term is used here, is work obtained, allocated, priced, monitored, and paid through a digital platform that stands between the worker and the customer and takes a share of each transaction. The platform is not a noticeboard. A noticeboard displays offers and leaves the parties to find each other, agree terms, and settle payment. The platform does the finding, sets or strongly steers the price, defines the standard the work must meet, routes the worker through the task, records the performance, handles the money, and can terminate the relationship by deactivation, which is exclusion from the market the platform operates. The worker supplies labour and, characteristically, some of the physical equipment the labour requires. The customer receives a service and pays the platform, not the worker. That triangular shape, worker, platform, customer, with the platform holding the connecting tissue, is the phenomenon. Everything else is variation.

The variations matter and will be treated separately in the platform class grid below, but the main forms can be named here. Ride hailing matches passengers to drivers for journeys priced by the platform’s formula. Food delivery matches orders to couriers for trips priced per drop. Domestic and care platforms match cleaners, carers, and childminders to households, often for recurring engagements. Online freelancing platforms match clients to designers, translators, programmers, and writers who may never leave their desks. Microtask platforms break digital work into fragments, image labelling, transcription snippets, content review, survey completion, paid in very small units and performed by workers scattered across continents. These forms differ in skill, in pay, in the equipment involved, and in how tightly the platform directs the work, and a serious analysis must let those differences show rather than dissolving them into a single portrait of the gig worker. The portrait habit is the first error of most commentary. There is no single gig worker. There are several distinct labour processes wearing a common contractual costume, and part of the value of the class grid is that it forces the differences into the open before any verdict is pronounced.

Several features recur across the forms and belong to the phenomenon rather than to any one sector. The work is paid by the task rather than by the hour in the great majority of arrangements, a payment form whose significance Marx analysed at length and which this article examines in its own section below. The worker’s access to work is mediated by an account that can be suspended, and the account accumulates a rating, a quantified reputation that belongs in practice to the platform’s system rather than to the worker, since it cannot be carried to a rival platform or converted into an independent business asset. The demand side is managed with prices that move, rising when demand surges and falling when it slackens, so that the worker’s earnings per task fluctuate with conditions the worker does not control. The costs of the equipment, fuel, insurance, maintenance, data, and the unpaid time spent waiting for allocation fall largely on the worker. And the relationship is framed, in the standard contractual documents, as one between independent businesses, a framing whose legal fortunes have varied and whose analytical significance is examined later. None of these features is universal in every detail, but together they describe the shape that any class analysis of the gig economy has to explain.

It is equally part of an honest description that the arrangement attracts many of its workers. Entry is fast, hours are in significant part chosen rather than assigned, the work can be combined with study, caring responsibilities, or another job, and there is no supervisor in the room. Workers who value those features are not deceived about them; the features are real. The strongest objection to the analysis offered in this article rests precisely on that reality, and it is given a section of its own at full strength below, because an analysis that cannot accommodate the genuine appeal of the arrangement to many of the people inside it is describing a different phenomenon from the one that exists. Description first, then categories, then objections, then verdict: that is the order this article keeps, and it is the order the template for application articles in this series requires.

Class in the Marxist scheme: the criterion before the case

Class analysis goes wrong most often by applying a criterion it has never stated. Three criteria circulate in ordinary usage and are constantly confused with one another. The first sorts people by income, so that class becomes a ladder of earnings with the working class at the bottom and the middle class in the middle. The second sorts by status and culture, so that class becomes a matter of accent, education, taste, and occupation prestige. The third, the Marxist criterion, sorts by relation to the means of production: who owns productive assets and lives by employing the labour of others, who owns no productive assets and lives by selling labour power, and who owns modest productive assets and lives principally by their own labour applied to them. The three criteria often overlap in practice, which is why the confusion survives, but they are not the same instrument, and they classify hard cases differently. A highly paid software contractor may stand high on the income ladder while standing, on the third criterion, exactly where a factory operative stands, since both sell labour power and neither employs others. A struggling shopkeeper may stand low on the income ladder while standing, on the third criterion, outside the working class, since the shopkeeper owns the shop and its stock. Anyone who wants to dispute the conclusions of this article is entitled to dispute the criterion, but the criterion must be disputed as a criterion, not bypassed by silently switching to income halfway through the argument.

What determines class position in Marxist analysis?

Class position is determined by a person’s relation to the means of production: whether they own productive assets and employ the labour of others, own none and must sell their labour power to live, or own modest assets and work with them themselves. Income, status, and self description may accompany a position, but none of them defines it.

Two consequences of the criterion are worth stating before the platform case is approached, because both do work later. The first is that class is a relational and structural notion, not a moral one. To classify a person as working class is not to praise or pity them, and to classify a person as petty bourgeois is not to accuse them of anything. The classification makes a claim about where a person stands in a system of production and about which interests that position tends to generate, and it is compatible with any view about the person’s character, politics, or satisfaction with their life. The second consequence is that class position and class consciousness are different things, a distinction the series treats in the class cluster and which matters acutely for platform work. A worker may occupy a working class position while understanding themselves as a business owner, a partner, or a free agent. That gap between position and self understanding is not, by itself, evidence against the classification, since the theory has always expected the forms in which a relation appears to influence how it is understood. But neither is the gap a licence to ignore what workers say about their own situation, and the objection section below takes the charge of condescension seriously rather than waving it away.

The criterion also has a boundary problem built into it, and platform work is not the first phenomenon to expose the seam. The classical scheme was drawn for an economy in which the typical producer either owned a workshop or owned nothing, and in which the means of production for most trades were substantial, indivisible, and expensive: a mill, a forge, a print works, a warehouse of stock. Where the means of production are cheap and divisible, a sewing machine, a set of tools, a laptop, the boundary blurs, because ownership of the equipment no longer separates those who can produce independently from those who cannot. Marx himself analysed trades in which workers owned their instruments while remaining dependent on merchants who controlled access to materials and markets, and the putting-out system of early industrial production was built on exactly that combination of worker-owned tools and merchant-controlled markets. The platform case is therefore not an anomaly the theory never contemplated. It is a modern form of a configuration the theory contemplated at its birth, which is one reason the analysis below leans so heavily on the question of which means of production are decisive. Where equipment is cheap, the decisive means migrate upward, from the tool to the system that connects the tool’s owner to the customer, and it is at that higher level that the class question must be asked.

The ownership puzzle: a worker who owns means of production

State the puzzle at its strongest. The proletarian of the classical definition is separated from the means of production: the mill owner owns the mill, the machinery, the raw material, and the product, and the operative owns none of these. The ride hailing driver, by contrast, owns the car, which is the single most expensive physical asset the labour process uses. The car is not a trivial possession. It is worth more than many small businesses’ entire equipment stock, it can be sold, it can be used for private purposes, and in some arrangements it can be hired out. The courier’s bicycle is worth less, but the structure of the point is identical: an asset without which the work cannot be done belongs to the worker, not to the company. If ownership of means of production is the criterion, and the worker owns a major means of production, then the classification seems forced: the driver is a petty bourgeois, a small proprietor who sells a service to the public and buys a matching service from the platform, and the whole language of employment, exploitation, and working class conditions is a category error dressed as solidarity. This is the argument the platforms themselves make in legal proceedings and in public advocacy, and it deserves to be met in its strong form rather than in the weak form, the suggestion that the car is somehow not real property, which nobody believes.

Are gig workers working class if they own their own vehicle?

Yes, on the decisive-means test. The vehicle is a cost shifted onto the worker, while the platform owns the matching infrastructure, the data, and the customer relationship, which are the means of production that actually govern the labour process. Ownership of a tool is not ownership of the process.

Notice what the puzzle presupposes. It treats means of production as a heap of physical objects and asks who holds title to the objects. But the classical definition was never really about title to objects in the abstract. It was about the conditions a producer needs in order to produce and to reach a market, and about whether those conditions are under the producer’s own control or must be accessed through an owner who can grant or refuse access. The handloom weaver owned the loom. What the weaver did not own, once the merchant system matured, was the yarn supply on credit, the designs the market wanted, or the route to the distant customer, and it was that dependence, not the loom, that determined the weaver’s fate and eventually the weaver’s class trajectory. The loom was real property and it availed its owner very little. Ownership puzzles of this shape are resolved not by counting assets but by asking which assets are decisive, which ones the labour process cannot proceed without and cannot substitute, and whose withdrawal would end the worker’s access to production altogether. A driver whose car is repossessed can rent another car and return to the road within days. A driver whose platform account is closed cannot reach the platform’s customers at all, on any car, however new. That asymmetry, between a replaceable asset and an irreplaceable gateway, is the analytical heart of this article, and the next three sections develop it into the three serious answers the literature and the case law offer to the ownership puzzle.

A further preliminary distinction will prevent confusion later. The question is not whether the driver is an employee. Employee is a legal status, defined by statute and case law, carrying specific rights, and it varies between jurisdictions and changes when legislatures intervene. The question here is whether the driver occupies a working class position in the relations of production, which is a claim about dependence, control, and the source of the income from which the platform lives. A person can be a non-employee in law and a worker in the analytical sense, as generations of homeworkers, outworkers, and day labourers were before employment law existed to name them. A person can also be an employee in law while occupying, in rare professional cases, a position whose income and autonomy make the classification genuinely contestable. The categories cross without coinciding, and keeping them apart is the single most useful discipline a reader can bring to public argument on this topic, where the two are merged in nearly every paragraph written on any side.

The first answer: the decisive means of production

The first and most fundamental answer holds that not all means of production are equal, and that class position is determined by ownership of the means that are decisive for the labour process in question. A means of production is decisive when the labour process cannot proceed without it, when the worker cannot substitute or reproduce it independently, and when control over it carries control over the terms on which the work is done: the price, the standard, the allocation, and the continuation of the relationship. By that standard, the decisive means in platform work are not the car or the bicycle. They are the matching and dispatch infrastructure that pairs worker with customer, the accumulated data about demand, routes, prices, and performance, the payment system, the rating archive, and above all the customer relationship, the fact that the passenger thinks of the platform, not of any driver, when a journey is needed. All of these the platform owns absolutely. No driver can build a rival dispatch network in a garage. No courier can carry a reputation score across to a competitor or take a regular customer’s future orders away as personal goodwill, because the orders were never addressed to the courier in the first place. Withdraw the car and the work pauses until another car is found. Withdraw the platform and the work, as this kind of work, ceases to exist.

Formulated as a rule, this is the decisive-means test, and it is the namable claim of this article: class position is determined by ownership of the means of production that are decisive for the labour process, not by ownership of any means of production, and applying that test places platform workers in the working class while explaining precisely why the question felt hard. The test is not an invention contrived for the platform case. It generalises what the classical analysis always did in practice. Nobody argued that a mill operative who owned a handkerchief, a lunch tin, and a set of personal tools was thereby a proprietor, because those possessions did not give access to production. Nobody argues now that a surgeon who owns a personal stethoscope owns the hospital. In every period, the analysis has silently distinguished the equipment a worker happens to hold from the productive system the worker must enter in order to work at all, and it has classified by the system. The platform case simply forces the silent distinction into the open, because the equipment held by the worker has grown large and expensive enough to be mistaken for the system. A car is a plausible candidate for decisive means in a way a lunch tin is not, which is why the puzzle is genuine, and why resolving it requires the test to be stated rather than assumed.

The test also discriminates, which is what a good test should do. Consider a plumber who owns tools and a van, holds a personal list of customers built over years, quotes prices directly, and takes payment directly. The plumber’s decisive means, the customer relationship and the pricing power that goes with it, are in the plumber’s own hands, and the classification follows: a petty bourgeois proprietor, however modest the turnover. Consider the same plumber working exclusively through a home services platform that assigns the jobs, sets the rates, forbids direct contact with customers, and can deactivate the account. The tools and the van have not changed. Everything decisive has moved to the platform, and the classification moves with it. Consider, at the other edge, a genuinely independent trader on an online marketplace who owns a recognisable brand, a repeat customer base that follows the brand rather than the marketplace, and the practical ability to move the whole operation to another channel. There the marketplace approaches a mere venue, the seller’s position approaches genuine proprietorship, and the analysis should say so rather than forcing the case into a mould. A test that classifies every platform participant identically would be a slogan. The decisive-means test classifies by where control of the labour process actually sits, and it therefore gives different answers for the marketplace brand owner and the ride hailing driver, which is a point in its favour rather than against it.

One refinement completes the answer. Decisiveness is a matter of the whole arrangement, not of any single asset considered alone. The platform’s dispatch system would be worthless without drivers and vehicles, just as the mill’s machinery was worthless without operatives, and interdependence of that kind settles nothing, because class analysis has never claimed that owners can produce without workers. The claim is about the direction of dependence at the point of access. The platform can replace any individual driver from a pool of applicants without interrupting its service for an hour. The individual driver cannot replace the platform’s accumulated demand, its mapping and pricing machinery, or its payment trust, at any price available to an individual. Asymmetry of replaceability is the practical fingerprint of the decisive means, and it is observable without any commitment to the rest of the theory: one party can walk away from the relationship and continue its business intact, and the other cannot. That observable asymmetry is what the remaining two answers approach from different directions.

The second answer: the vehicle as a cost shifted onto the worker

The second answer examines the owned asset itself and asks what economic function it performs. In the classical factory, the capitalist advances constant capital: buildings, machinery, materials, fuel. The cost of that equipment, including its wear, is recovered from the product, and its ownership is the material basis of the capitalist’s claim on the surplus. In platform work, a striking share of that advance has been transferred to the worker. The driver finances the car, insures it, fuels it, maintains it, and absorbs its depreciation, including the accelerated depreciation that intensive commercial use produces. The courier supplies the bicycle, the phone, the data contract, the waterproof clothing, and the body that wears out in traffic. The cleaner supplies the materials. None of these costs appears in the platform’s accounts as capital advanced. All of them are real costs of the production process, without which no journey, delivery, or clean would occur. The arrangement is therefore better described as a deduction structure than as proprietorship: the worker’s gross receipts must first cover the costs of the equipment and its upkeep before anything resembling a wage remains, which makes the vehicle functionally closer to a compulsory purchase the worker must make in order to access work than to an asset conferring an owner’s position. An asset whose main economic role is to absorb costs that a conventional employer would bear, and whose owner cannot use it to employ anyone else or to bargain over the price of the final service, is a peculiar kind of means of production. It produces obligations more reliably than it produces independence.

The series treatment of surplus value in modern firms works through the accounting side of this point, showing why published accounts cannot settle questions about where value is produced and captured. The class significance can be put in terms of risk. Ownership in the capitalist sense is normally accompanied by the powers that make risk worth bearing: the power to set the price, to choose the product, to expand, to hire, to sell the business as a going concern. The platform worker bears a proprietor’s share of the risk, the vehicle loan that persists through a slack month, the repair bill that arrives without warning, the earnings that fall when the platform changes its formula, while holding almost none of a proprietor’s powers. The price is set by the platform. The product, the standard journey or delivery with its time targets and conduct rules, is specified by the platform. Expansion by hiring another driver to work one’s account is generally prohibited by the terms of service. The business cannot be sold, because the account and its rating are non-transferable. What remains, when the powers are subtracted and the risks are retained, is not the petty bourgeois position of the classical scheme, in which the small proprietor at least commands the little operation that bears the risk. It is a hybrid in which the cost bearing has been individualised and the command has been centralised, and class analysis should name the hybrid accurately rather than accepting the half of it that appears in the contract.

There is a historical rhyme here that the tradition supplies ready made. In the domestic system and the early factory districts, merchants and manufacturers repeatedly shifted equipment costs onto workers: frames rented to stockingers at charges deducted from earnings, tools purchased from the employer, workshop rent and power charges stopped out of the piece rate. Contemporary critics and later historians treated those deductions as part of the wage relation’s machinery, not as evidence that the stockinger had become a capitalist. The truck system and the frame rent scandals were objectionable precisely because they made the worker pay for the privilege of working, and legislation eventually restricted the practice in several industries. The platform arrangement differs in technology and in legal form, but the cost shifting function is recognisably the same, and the analytical conclusion should be the same as well. When the worker’s asset is acquired under the pressure of needing access to work, maintained out of task earnings, and useless as a lever over price or conditions, its ownership is a feature of how the wage relation is organised in this sector, not a refutation of the wage relation’s existence.

None of this requires denying that the vehicle is genuinely the worker’s property, or that some workers use platform earnings to build real assets, or that a minority graduate into operating small fleets and employing others, at which point their class position genuinely changes and the analysis should register the change. Class positions are occupied by people in motion, and a test is not refuted by the existence of people crossing the boundary it draws. The second answer’s claim is narrower: that the typical platform worker’s equipment ownership functions as shifted constant capital and as a barrier the worker must finance, and that counting it as proprietorship without examining its function mistakes the legal form of an asset for its economic role. Combined with the first answer, the picture sharpens. The platform owns what is decisive. The worker owns what is costly. The contract calls the second fact independence and passes over the first in silence.

The third answer: disguised employment and the legal route

The third answer comes not from theory but from litigation, and it holds that the arrangement is disguised employment: that the platform worker is, in substance, a worker working for another’s business under that business’s direction, and that the contractual language of independence describes the substance inaccurately. Courts and tribunals in multiple jurisdictions have been asked to decide versions of this question, and although their conclusions differ and continue to shift as legislatures intervene, the reasoning they share is instructive for the analytical question, because the legal tests courts apply track, in a rough juridical dialect, the same features the class analysis examines. The standard tests ask who controls the manner of the work, whether the worker must perform it personally or may send a substitute, whether the worker is integrated into the business or running a distinct business alongside it, who bears the financial risk, and whether the written terms match the reality of the arrangement as actually practised. Where adjudicators have looked past documents to practice, they have repeatedly found platform terms, unilateral price setting, performance monitoring, discipline through ratings and deactivation, restrictions on building an independent clientele, that sit awkwardly with the picture of two businesses contracting at arm’s length. Where legislatures have intervened, some have created or confirmed intermediate categories, a worker status between employee and self employed, and some have moved in the opposite direction, writing platform specific contractor status into law, in at least one large subnational jurisdiction by ballot measure after an intense public campaign. The landscape is contested and unsettled, and this article describes it as such rather than as a settled legal position, because it is not settled and because the analytical argument does not need it to be.

What do the legal tests for employment status actually examine?

Legal tests examine control over how work is done, whether personal service is required or a substitute may be sent, integration into the business, who bears financial risk, and whether written terms match actual practice. They ask whether the worker runs an independent business or works inside another’s business under its direction.

The analytical value of the legal route is double edged, and both edges should be acknowledged. On one side, the litigation record is a rich empirical source. Cases compel disclosure of how allocation, pricing, and discipline actually operate, and the factual findings of tribunals, the descriptions of acceptance rates, of information withheld from workers about destinations and fares, of penalties attached to refusal, constitute some of the most reliable evidence available about the interior of the platform labour process. A class analysis that ignored that record would be choosing purity over evidence. On the other side, the legal question is not the analytical question, and the third answer is the weakest of the three when it stands alone. Legal categories are instruments for distributing specific rights and obligations, drafted for administrability, riddled with jurisdictional accident, and revisable by any legislature that dislikes a court’s conclusion. A classification that a ballot measure can reverse is a classification about law, and a class analysis that simply borrowed it would inherit every one of those contingencies. The durable use of the legal material is corroborative. When courts applying functional tests, and theorists applying the decisive-means test, and workers describing their own experience, converge on the same features, unilateral price setting, dependence on the platform’s gateway, discipline without a human supervisor, the convergence is evidence that those features are the real joints of the arrangement. The third answer therefore supports the first rather than replacing it. The platform worker is not working class because a tribunal said so. The tribunal and the analysis are responding to the same underlying structure, each in its own vocabulary.

There is also a caution in the legal material that cuts against careless use by the analysis’s own side. Some platform arrangements survive functional scrutiny better than others. Genuinely open marketplaces, where sellers set prices, build followings, and communicate freely with customers, present a different structure from dispatch platforms, and legal disputes about them have correspondingly different shapes. Skilled freelancers who negotiate rates across several platforms and direct clients, maintain waiting lists, and refuse work at will occupy a position in which several of the control tests point the other way. The third answer, honestly used, therefore reinforces the discriminating character of the whole analysis. The question is never the label on the contract and never the sector in the abstract. It is always the configuration of control, dependence, and access in the specific arrangement, which is exactly what the decisive-means test formalises.

Formal and real subsumption: the sharpest instrument

The three answers classify the platform worker. A distinction drawn from Marx’s unpublished manuscript material, the chapter on results of the immediate process of production that was drafted for the first volume of Capital and published only later, explains the mechanism by which the platform rules, and it is the sharpest instrument this analysis possesses. Marx distinguished two ways capital takes command of labour. Under formal subsumption, capital takes over a labour process it finds already existing and leaves its technical character largely unchanged, while transforming the social relation around it: the producer who once sold a product now sells labour power, works under the capitalist’s supervision, and produces a surplus the capitalist appropriates. The handloom weaver working for a merchant on materials the merchant supplies is formally subsumed; the weaving is the same weaving, performed with the same tools in the same cottage, but the weaver’s product and time now belong to another. Under real subsumption, capital goes further and revolutionises the labour process itself, through cooperation, the division of labour in manufacture, and finally machinery and large scale industry, so that the way the work is technically done, its pace, its sequence, its tools, its very motions, is designed by capital for the extraction of relative surplus value. The factory operative is really subsumed; no pre-capitalist version of that job existed, because the job is an artefact of the system that employs it.

What is the difference between formal and real subsumption?

Formal subsumption is capital taking command of a labour process it finds ready made, leaving the technique of the work largely intact while changing who appropriates the surplus. Real subsumption is capital remaking the labour process itself, redesigning tasks, tools, sequence, and pace so that the work could not exist outside the capital relation that organises it.

Platform work is fascinating to the analyst because it appears, at first sight, to be the purest modern specimen of formal subsumption, and in one respect a regression behind even the factory’s beginning. The driver drives, as drivers have always driven, in a vehicle the platform does not own, on roads the platform does not maintain, using skills learned outside the platform’s employment. The courier rides a bicycle, a technology older than the light bulb, through streets the courier often knows better than the application does. No machinery of the platform’s stands over the worker; there is not even a workplace. Capital seems merely to coordinate, to stand at the gate of a pre-existing activity and take a toll for arranging introductions, and that is exactly how the platforms describe themselves, as technology companies providing a marketplace rather than as employers organising production. If formal subsumption means capital coordinating a labour process it has not transformed, the platform presents itself as formal subsumption reduced to its minimum, coordination with almost nothing else, and the toll begins to look like rent for access rather than the appropriation of surplus value in production. That appearance is the strongest theoretical version of the platform’s own self description, and it is why the subsumption analysis, and not only the ownership analysis, is required to settle the question.

The appearance does not survive acquaintance with the labour process as actually performed. Consider what the platform’s system does to a journey. It decides which driver is offered which passenger, and in what order, according to criteria the driver cannot inspect. It sets the route the driver is expected to follow and registers deviation. It determines the price of the journey by a formula that varies with conditions and that the driver learns only after accepting, and in many arrangements the destination itself is withheld until acceptance, so that the worker commits to labour before knowing its object. It times the stages of the work, the arrival, the wait, the trip, and it records customer judgement in a score that conditions future access to work, so that discipline is continuous, quantified, and automatic. It manages the pace of the whole system by adjusting incentives, bonuses for sequences of trips, penalties in allocation for refusal, surge prices that summon labour to where it is wanted with an efficiency no foreman could match. Each of these is an intervention in the labour process itself, in its allocation, sequence, pace, information structure, and standards, and taken together they amount to a redesign of the work of driving for hire as thorough as the redesign of weaving that the factory accomplished, achieved without owning a single vehicle or building a single depot. The labour process has been transformed, not merely coordinated. What the platform coordinates is the market. What it subsumes, really and not merely formally, is the work.

The subsumption inversion

That finding deserves to be stated as the inversion it is, because it reverses the expectation the classical sequence creates. In the standard narrative, real subsumption arrives wearing machinery: capital first takes over the process formally, then rebuilds it around powered equipment it owns, and ownership of the transformed process and transformation of the process advance together. Platform capital separates the two. It achieves real subsumption of the labour process while deliberately refusing the ownership of the physical instruments that the classical sequence bundled with it, because the instruments are precisely the costs it wishes the worker to carry. The transformation is informational rather than mechanical. Where the factory redesigned work by rearranging bodies around a power source, the platform redesigns work by rearranging information around a worker who supplies the body and the vehicle alike: what the worker knows, when the worker knows it, which task arrives next, what it pays, how it will be judged. Control over information is control over the labour process when the labour process consists of moving through a city responding to demand, and the platform’s monopoly of the information layer is the exact counterpart of the mill owner’s monopoly of the powered machinery. The decisive means identified by the ownership analysis and the transforming power identified by the subsumption analysis turn out to be the same thing viewed from two sides, the dispatch and data infrastructure, which is owned absolutely and which does the reorganising. The two analyses converge, and convergence between independent lines of argument is the strongest warrant a classification can have.

The inversion also explains a feature of platform work that puzzles commentators who approach it with factory expectations, namely the coexistence of intense control with the absence of visible supervision. Real subsumption in the factory required the overseer because the machinery set the pace but could not allocate tasks, judge quality, or dismiss. The platform’s system performs all four functions continuously and impersonally, and the worker experiences the result as a stream of offers, ratings, and notifications rather than as orders. Experienced as freedom from bosses, the arrangement is nonetheless experienced also, by many workers, as a peculiar opacity: decisions that govern earnings arrive without reasons, standards shift without announcement, and the appeal against an automated judgement leads to another automated judgement. The vocabulary of the cybertariat, one of the coinages circulating for this workforce, and the older vocabulary of algorithmic management, name the same configuration from different angles. What the subsumption analysis adds is the explanation of why measurement matters. Algorithmic direction is not merely supervision with better instruments. Because every stage of the work is recorded, the labour process becomes measurable to a degree the factory never achieved, and measurability is the precondition of intensification, of squeezing more labour into the same hour by closing the gaps, the idle minutes, the unrecorded pauses, that older forms of supervision could not reach. The gaps in the working day are where the platform’s redesign bites hardest, in the waiting time it declines to pay for, the repositioning it directs but does not remunerate, the sequence discipline that keeps the worker continuously available. An analysis that classified algorithmic management as a new style of supervision would miss the quantitative change inside the qualitative one, which is why the brief for this article lists that misreading among those to be corrected.

A boundary should be marked before the argument is carried further, because the subsumption analysis, like the ownership analysis, discriminates rather than sweeping everything before it. The degree of real subsumption varies across platform types in step with the degree of informational control. A dispatch platform that withholds destinations, sets prices, and allocates by algorithm subsumes really and deeply. A listing platform on which independent professionals publish portfolios, negotiate scope and price with clients directly, and use the platform chiefly for payment processing and discovery subsumes weakly and mostly formally, taking command of the transaction while leaving the labour process, the design work, the writing, the consulting, substantially in the producer’s own hands. Between those poles lie most real arrangements, and the analysis earns its credibility by placing each type on the scale rather than by pronouncing a single verdict on a heterogeneous sector. The class grid that follows is built to do exactly that placement work, row by row, with the three answers to the ownership puzzle applied to each type in turn.

The rating: an asset the worker produces and the platform owns

One element of the platform arrangement condenses the whole analysis into a single object, and it deserves examination before the grid assembles the parts: the rating. A driver’s or courier’s score is produced by labour. It accumulates out of hundreds of journeys completed on time, of courtesy maintained under provocation, of vehicles kept clean and routes learned and customers soothed, and it represents, in quantified form, the worker’s demonstrated reliability, which in an older economy would have been the worker’s most personal professional asset, a reputation. The character reference of the domestic servant, the standing of the journeyman with the masters of a town, the goodwill of the tradesperson’s name over the door: each was a reputation the worker carried from engagement to engagement, imperfectly and sometimes unjustly, but as a possession that travelled with its owner and could be spent in a new place. The rating travels nowhere. It is recorded in the platform’s system, computed by the platform’s formula, weighted by factors the worker cannot inspect, and extinguished without residue if the account closes. A worker with a decade of flawless service who leaves a platform, or is removed from one, starts at a rival platform indistinguishable from a novice, because the proof of the decade belongs to the company that recorded it. Reputation, which the independent producer owns as goodwill and can even sell with a business, has been separated from the producer and attached to the gateway, where it functions as one more component of the decisive means: the platform owns not only the connection to customers in general but the worker’s own certified history with customers in particular.

The parallel with the article’s central distinction is exact enough to serve as a test of it. Ask why the non-portable rating matters, and the answer reproduces the decisive-means logic in miniature. The worker can replace the car, the bicycle, the phone, and every physical instrument of the work, but cannot replace the accumulated score, because the score’s value consists entirely in its recognition by the system that allocates work, and recognition cannot be transferred by the person it describes. The rating is therefore a means of production in the strictest functional sense: a produced condition of further production, without which the worker’s access to work degrades to the bottom of the allocation queue. That it is intangible is irrelevant; the dispatch data and the customer network are intangible too, and the analysis has already established that the decisive means in this sector are informational. What the rating adds is biographical poignancy. Every other asset the platform owns was built by the platform’s engineers and investors. The rating archive was built by the workers, trip by trip, and is owned by the platform as completely as the servers it sits on. The product of the workers’ collective labour, their certified reliability, confronts each worker as the property of another and as a condition of continued employment, which is a structure the classical texts described in the context of machinery and which reappears here in the unlikely form of a number between one and five.

Campaigns for rating portability, which surface wherever platform work is regulated, understand this intuitively. To make the score portable would be to return one component of the decisive means to the workers, and the resistance such proposals meet, on grounds of data standards and competitive confidentiality, measures the component’s value to its owner. Portability would not dissolve the platform’s position, since dispatch, pricing, and demand would remain, but it would change the exit calculation for every experienced worker, and exit calculations are the substance of dependence. The point will return in the section on what would genuinely change the classification, where rating portability appears as one of the few reforms that touches the structure rather than the symptoms.

The platform class grid

The findable artifact of this article gathers the argument into a single reference table. Five platform work types are set against what the worker owns, what the platform owns, the degree of algorithmic control exercised over the labour process, the pattern of legal classification that disputes about that type have produced, and the resulting class position. The classification column compresses the three answers, decisive means, shifted costs, and the legal corroboration, into the verdict they jointly support for each row, and the prose following the table unpacks the rows whose verdicts need defence.

Platform work type	What the worker owns	What the platform owns	Algorithmic control	Legal classification pattern	Class position under the three answers
Ride hailing	The vehicle, phone, fuel and running costs	Dispatch and matching system, pricing formula, passenger data, payment system, brand	Very high: allocation, pricing, routing, ratings, deactivation	Contested; functional tests have often found worker status, some legislatures have written contractor status instead	Working class: decisive means and price setting sit wholly with the platform
Food delivery	Bicycle or vehicle, phone, bag, protective clothing	Order allocation, merchant and customer relationships, fee structure, delivery data	High: batching, time estimates, acceptance metrics, ratings	Contested; rider specific presumptions of employment in some systems, contractor status in others	Working class: the classic shifted-cost case on the smallest capital
Domestic and care work	Cleaning materials, transport, sometimes specialist equipment	Client matching, vetting archive, booking and payment system, review record	Moderate: matching and reputation strong, direction of the work itself weaker inside the home	Frequently treated as self employment; agency and employee models coexist and disputes turn on the specific scheme	Working class where the platform controls access and price; nearer petty bourgeois where the worker builds a direct clientele
Online freelancing	Computer, software, skills, portfolio	Marketplace visibility, ranking and search placement, payment escrow, dispute process	Low to moderate: ranking governs access, the labour process itself mostly self directed	Generally treated as independent contracting, with disputes centred on fees and account closure rather than status	Mixed: petty bourgeois where reputation and clients are portable, working class in tendency where ranking dependence is total
Microtask work	Computer or phone, connectivity, electricity	Task supply, qualification tests, quality scoring, payment system	High in allocation and quality control, though the tasks are fragments rather than a continuous process	Rarely litigated as employment; usually classified as casual contracting across borders	Working class: no asset, no pricing power, and no route to the requesters except through the platform

Read the grid by its contrasts. The first two rows are the strong cases, where every instrument of the analysis points the same way, and they are also the rows that dominate the public image of the gig economy. The third row is the instructive middle, because the labour process inside a private home resists the platform’s direction in ways driving does not: the platform can allocate the cleaner and archive the rating, but the cleaning itself is specified by the household, performed unwatched, and repeated for regular clients whose loyalty may attach to the person rather than to the application. Where that loyalty becomes direct engagement, the worker begins to migrate across the boundary the grid draws, and the migration is observable in the common practice of platforms attempting to prohibit off platform contact, a prohibition that exists precisely because the decisive means in this row, the client relationship, is contestable in a way the ride hailing relationship is not. The fourth row spreads widest. A translator with a portable reputation, direct clients on three continents, and a platform account kept as one channel among several is a petty bourgeois professional in everything but income volatility, while a designer whose entire order flow arrives through a single marketplace ranking, whose prices are disciplined by that ranking’s visibility rules, and whose account closure would mean starting from nothing, stands in a dependence structurally similar to the driver’s, softened by skill scarcity. The fifth row is the limiting case in the opposite direction: the microtask worker owns almost nothing, not even a vehicle to generate the puzzle, and the arrangement shows what the other rows look like when the shifted asset is subtracted, which is a wage relation in fragments, paid by the piece, supervised by software, and denuded of every incident of employment. The grid’s lesson is that the gig economy is not one class location but a family of them arranged along the variables the analysis has isolated, control of the decisive means, depth of subsumption, and portability of the customer relationship, and that the family’s centre of gravity lies firmly on the working class side of the line.

The third side of the triangle: the customer as inspector

Class analysis of a two party relation, worker and platform, omits the party whose behaviour disciplines both, and the omission should be repaired before the rows are examined one by one. The customer in platform work is not a passive recipient. Through the rating, the customer is enrolled as an inspector of the labour process, supplying the judgements that the platform’s disciplinary system consumes. The arrangement is an elegant solution to the oldest problem of service management, that the work is performed away from the supervisor’s eye, in the customer’s car journey, at the customer’s door, inside the customer’s home. Rather than supervise the worker directly, the platform delegates supervision to the person already present, arms that person with a scoring instrument, and aggregates the scores into consequences the customer neither sees nor intends: the allocation drought, the warning, the deactivation threshold approached one four star journey at a time. Customers mostly experience the instrument as a courtesy, a chance to reward good service, and the asymmetry built into the scale, on which anything below the maximum counts against the worker, is invisible from the customer’s side of the screen. The worker experiences the same instrument as a continuous plebiscite conducted by strangers, and adjusts the labour accordingly, performing the emotional work of pleasing as part of the job’s real content while its formal description mentions only transport or delivery.

The triangle also has a price side. The customer’s convenience, the car in three minutes, the meal tracked to the door, is produced by the worker’s unpaid availability, the circling, the waiting, the readiness that the per task payment refuses to count. Consumers of platform services therefore enjoy a standard of service whose cost is partly borne by labour they never see, in intervals the system defines out of the working day, and the political economy of the sector’s popularity cannot be understood without that transfer. Cheapness and immediacy are not properties of the software. They are properties of a labour arrangement that keeps a reserve of workers in unpaid readiness around the city, and the demand side’s attachment to the arrangement, the genuine delight of the customer, is one of the forces that stabilises it politically, aligning millions of small consumer interests against the claims of a workforce that remains, to most customers, a sequence of interchangeable arrivals. A class analysis that ignored the customer would miss why the sector’s controversies so often stall: the third side of the triangle votes, consumes, and experiences itself as uninvolved, while its ratings move the allocation and its expectations set the tempo the other two sides must meet.

Ride hailing: the fullest case

Ride hailing is the fullest case because it combines the largest worker owned asset with the most complete platform control, and the combination is what makes it analytically precious. The older trade it displaced is instructive. The taxi driver of the regulated city was never a simple figure either: some owned their vehicles and licences and were genuine small proprietors, some rented vehicles from fleet owners by the shift in an arrangement contemporaries described in the language of exploitation, and some were employees pure and simple. The licence, the medallion where that system operated, was often the decisive asset, scarce by regulation and valuable enough to be traded and financed. The platform’s innovation was to dissolve the licence as the gateway and substitute its own gateway, the application and its accumulated demand, which it owns without regulatory scarcity and expands without limit. The driver who under the old system might have aspired to own the decisive asset, saving toward a licence, now faces a decisive asset that cannot be bought at any price an individual can pay, because it consists of the network itself. The aspiration structure of the trade changed with the technology, from a ladder, however steep, toward a gate that does not open.

Inside the labour process, the platform’s direction exceeds anything the taxi dispatcher achieved by radio. Allocation is algorithmic and opaque; the driver learns to read the system’s moods, the surge areas, the airport queues, the hours when acceptance discipline is rewarded, and develops a practical science of an employer that cannot be questioned. Knowledge of that kind, the workforce’s accumulated lore about how the allocator behaves, is itself evidence of real subsumption, since workers develop such lore only where a designed system governs the work in detail. The rating regime completes the picture by deputising the passenger as an inspector whose judgements the driver must manage with the emotional labour of courtesy performed under time pressure, a requirement written into no contract and enforced by arithmetic. When the analysis says the ride hailing driver is working class, it is summarising all of this: an expensive asset owned and financed, a labour process directed in its allocation, pricing, routing, and standards by the owner of the decisive means, a product, the priced journey, whose price the producer never sets, and an exit that leaves the producer with a car and without a market. The classification is not a metaphor and not a gesture of sympathy. It is the output of the same tests that classify the plumber with a personal clientele the other way.

Food delivery and the courier’s bicycle

Food delivery strips the puzzle to its minimum, and for that reason it is where public argument most often begins. The courier’s owned means of production may amount to a bicycle worth less than a month of the platform’s marketing expenditure in the same city, and yet the contractual architecture is identical to the driver’s, independent business, own equipment, freedom to log on and off. If the bicycle courier is a petty bourgeois, the category has been emptied of content, since every worker in history who owned a tool would qualify, including the journeymen whose struggles fill the early chapters of labour history. The delivery case also displays the shifted cost structure in its most legible form, because the gap between gross receipts and net earnings is dominated by items a conventional employer supplies without discussion: the vehicle and its repair, the phone and its contract, the clothing that makes winter work survivable, and the unpaid intervals, waiting at restaurants for orders that are not ready, waiting between orders in zones the application suggests, which the per drop payment renders invisible. The industry’s own phrase for the worker’s productive time, time on task, performs the erasure by definition, counting as work only the minutes the platform chooses to remunerate and treating the availability that makes those minutes possible as the worker’s private affair. Marx’s analysis of the working day was built on the boundary struggle over what counts as labour time, and the delivery case shows the same struggle conducted by definitional means, inside an application, without a factory bell.

The courier case also contributes the clearest view of collective action in the sector, which the class consciousness discussion below takes up: wildcat stoppages coordinated through messaging groups, log off strikes timed to peak demand, and campaigns over per drop fees have recurred in city after city, organised by workers whom the contractual form describes as separate businesses and whom the work process isolates from one another almost completely. That such action emerges at all, under those conditions, and that it takes the classical forms of the strike and the fee demand rather than the forms of small business association, trade lobbying, price coordination among proprietors, is a datum the analysis is entitled to weigh. People classified against their real position can act from their real position, and couriers in dispute behave as workers disputing a wage, because the fee per drop is a wage in every respect except the name.

Domestic and care work through platforms

Domestic and care platforms deserve separate treatment because they carry the analysis into the home, where the traditions of the subject are oldest and the measurement problems hardest. Cleaning, childcare, elder care, and household help were performed for wages long before they were performed in factories, and they were performed for centuries under relations, service in a household, that class analysis has always found resistant, poised between employment, dependence, and a familial idiom that disguised both. The platform enters this old terrain as an intermediary that standardises discovery, vetting, booking, and payment while disclaiming the employments that surround it, and the disclaimer is more plausible here than in transport, because the platform genuinely does not direct the scrubbing or the feeding in detail. The household does, or the work directs itself by its nature. What the platform controls is access and reputation: the archive of reviews that makes a cleaner bookable, the background check that makes a carer acceptable, the payment channel that makes the arrangement feel safe to both sides. Where the platform also steers prices, through suggested rates, visibility rankings that reward lower prices, and fees taken from both sides, the decisive means analysis begins to bite; where the worker converts platform introductions into a direct book of regular clients, it relaxes. The row in the grid is therefore genuinely double, and honest analysis leaves it double rather than tidying it.

Two further features distinguish this row. The workforce is heavily female and, in many cities, heavily migrant, which connects the platform form to the older history of domestic service as a sector where the wage relation wore personal dress, and where the boundary between paid work and unpaid obligation was policed by custom rather than contract. And the work is care, whose product is partly the relationship itself, which resists the intensification that measurability enables elsewhere: a platform can time a clean, but the quality of attention given to an elderly person does not decompose into tasks without remainder, and attempts to make it do so generate the sector’s characteristic discontents. The class analysis of care platforms must therefore borrow from the series treatment of care work as well as from the class criterion, and its verdict, working class in position for the platform dependent cleaner and carer, rests less on subsumption of the labour process, which is shallow here, than on the first answer’s ground, dependence on a gateway the worker does not own for access to the households that are the market.

Online freelancing and microtask work: the dispersed cases

Online freelancing and microtask work mark the two extremes of dispersion, work performed anywhere, for requesters anywhere, by workers who may never learn one another’s names. The freelance marketplaces host the gig economy’s most skilled participants, and they are the strongest exhibit for the other side of the classification argument. A programmer or translator with scarce skills confronts the platform from a position of relative strength: the ranking matters, but the skill is portable, the portfolio can be shown elsewhere, direct relationships with clients are common and sometimes encouraged by the structure of long projects, and the fees, though resented, are closer to a market access charge than to the toll on a captive. The honest verdict for the top of this row is that many such freelancers are what the classical scheme says they appear to be, independent small producers selling services in a market, petty bourgeois in position whatever their income, and that the platform’s own description fits them tolerably well. The same marketplace’s lower reaches tell another story, of entrants in low wage countries competing for commodified tasks at prices set by global oversupply, of visibility algorithms that function as allocation, of account metrics that function as discipline. Skill scarcity is the variable that moves a freelancer along the row, because scarcity is bargaining power, and bargaining power is what the decisive-means test detects from the other side: the scarce professional can replace the platform, and the replaceability asymmetry runs, unusually, in the worker’s favour.

A further pressure works on the freelancing row from inside the marketplace’s own development, and it pushes the row’s centre of gravity toward dependence over time. Marketplaces that begin by charging a commission on completed contracts discover that visibility itself can be sold: promoted placement in search results, featured profiles, and paid access to the better class of job postings convert the worker’s need to be found into a second revenue stream drawn from the workers rather than from the clients. The purchase of visibility is rent in the strictest shape, payment for position within a scarce ordering the platform owns and manufactures, and the freelancer who must buy back the visibility that a reputation once supplied unaided is experiencing, in a mild but exact form, the separation of reputation from producer that the rating analysis described. Fees, meanwhile, are revised by the party that owns the gateway, and the history of mature marketplaces is a history of such revisions absorbed by sellers who cannot coordinate a refusal. None of this abolishes the genuine independence of the scarce professional at the row’s top. It explains why the row does not stand still, and why a classification performed once, on the marketplace’s younger and simpler form, would decay if it were not reperformed against the arrangement as it hardens.

Microtask work has no such upper reach. The tasks are defined so that no individual worker accumulates scarcity; the qualification tests, the quality scores, and the payment thresholds are all platform property; the requesters are anonymous and unreachable; and the work is fragmented precisely so that no labour process, with a beginning and an end that a worker could survey and improve, ever forms. If the factory divided labour to make each operative’s contribution partial, microtask platforms divide it to make each contribution infinitesimal, and the wage form follows suit, payments denominated in cents for units of attention. The workforce is global in a way the factory never was, which produces the sector’s defining political fact: the reserve of available workers is planetary, visible to the requester as an inexhaustible queue, and invisible to itself as a collectivity. Class analysis has a name for the condition of a workforce whose members compete without meeting, and the name is not a new class. It is the working class in its most atomised historical form, resembling the outworkers of the putting-out system more than the factory hands of the classical texts, a resemblance the next section develops.

Piece work returns: the wage form that hides the wage

The gig economy is routinely described as a return to piece work, and the description is correct in a sense deeper than the journalists who use it usually intend. Marx devoted a chapter of the first volume of Capital to piece wages, arguing that they are a converted form of time wages rather than an alternative to them, and that the piece form is the form of wages most adequate to the capitalist mode of production. The argument deserves reconstruction, because it explains the platform payment system’s effects with a precision that commentary on incentives rarely matches. A piece rate appears to pay for a product rather than for labour power, and the appearance transforms the worker’s understanding of the exchange: the courier is paid for the delivery, not for the hour, and therefore experiences waiting time, travel to the zone, and the slow restaurant as personal misfortunes rather than as unpaid portions of a working day. The rate itself is set so that the average worker, working at the intensity the system requires, produces earnings that correspond to the value of labour power for the period, with the correspondence concealed by the unit. Intensity is then enforced by the payment form itself, since the worker who pauses earns nothing, and supervision can be relaxed precisely because the wage form does the supervising. Marx noted the consequences that follow in train: the worker’s interest in lengthening the day and intensifying the hour without any overseer’s compulsion, the facilitation of parasitic intermediaries who take a cut between the principal and the actual worker, the use of the piece form to depress the earnings of the slower below subsistence while advertising the earnings of the fastest as the norm, and the way individual competition between workers replaces collective standards with personal records.

Why did Marx treat piece wages as the wage form most adequate to capitalist production?

Piece wages pay for the unit of output rather than the hour, so the wage relation disguises itself as a sale of products. The rate is calibrated to yield the value of labour power at the required intensity, supervision becomes unnecessary because stopping means earning nothing, and workers compete against one another’s records instead of maintaining a collective standard.

Platform payment is piece payment perfected by software. The per trip and per drop rates are adjusted dynamically, which the nineteenth century piece master, bound to a printed list, could not do, and the adjustment serves the classical functions simultaneously: it summons labour when demand rises, dismisses it without cost when demand falls, and calibrates the earnings of the median worker to the level the local labour market requires, while the surge episodes supply the advertising figures. The intermediary function also returns in digital dress, since between the customer and the worker stands a party that produces neither the journey nor the meal’s movement and yet takes a share of every unit sold, a share defended as a technology fee in the same vocabulary that defended the middleman’s cut in the sweating system. To call the gig economy a return to piece work is therefore not a metaphor about irregularity. It is an exact classification of its wage form, and it carries the classical analysis with it: the piece form presupposes the wage relation it disguises, because a rate calibrated to the value of labour power over a period is a time wage wearing the costume of a price. The worker who sells deliveries by the unit and the worker who sells hours by the week stand in the same relation to the buyer of their labour. Only the bookkeeping differs, and bookkeeping, however sophisticated, is not a class boundary.

The return of the intermediary: fleets, agencies and layers

Marx’s piece wage chapter contains a secondary observation that platform developments have made vivid: piece payment facilitates the interposition of parasitic intermediaries between the principal and the worker, layers that take a portion of the piece price for arranging access to the work. The sweating system of the garment trades ran on such layers, contractor upon subcontractor, each clipping the rate, until the final worker’s earnings bore little relation to the price the first buyer paid. Platform work was supposed, in its founding rhetoric, to abolish intermediaries by connecting parties directly. In practice the layers have re-formed inside the platform form, and their anatomy is instructive. In several markets, drivers work not for the platform directly but for fleet partners who own vehicles and accounts in bulk, renting cars to drivers or engaging drivers to work accounts the partner controls, and taking a share of the earnings for the arrangement. Around food delivery, agencies in some cities supply couriers to platforms or operate the local logistics under contract, standing between the application and the rider as employer in fact while the application’s terms describe a marketplace. Around online work, intermediary agencies assemble teams of freelancers under a single storefront, allocating tasks internally and retaining a margin invisible to both client and worker. Each layer reproduces the classical structure: the intermediary produces none of the service, owns a position between the producer and the gateway, and lives from a deduction. The layers matter for classification because they show the wage relation’s resilience under relabelling. Strip the platform’s own contract away and insert a fleet, and the driver does not become more independent; the dependence acquires a second storey. The analysis that looked only at the platform’s documents would miss the fleet entirely, which is why the decisive-means test is framed around the labour process as practised rather than around the contracting chain as drawn.

Incentives and the managed week: bonuses, streaks and the price of the next hour

Between the piece rate and the worker stands a second instrument that commentary treats as decoration and the analysis should treat as wage policy: the incentive system. Platforms deploy a shifting architecture of bonuses for completing sequences of tasks, premiums for working defined peak windows, guarantees conditional on acceptance rates, and streak rewards that reset if the worker pauses. Each element is a device for shaping the supply of labour hour by hour without issuing anything so crude as an instruction, and together they compose the closest thing the sector has to a rota, written in prices rather than in orders. The streak deserves particular attention, because it manufactures its own compulsion: a worker four trips into a five trip bonus experiences the fifth trip as already partly earned, and the decision to stop working is reframed, by the architecture of the reward, as the decision to surrender money in hand. Behavioural design of this kind is the piece wage system’s intensification function refined by experiment, since the platform can test incentive shapes against millions of decisions and retain the shapes that move labour most cheaply. Tips belong in the same analysis, whatever their surface warmth. Where tipping is built into the payment flow, a portion of the worker’s expected earnings is transferred from the platform’s rate to the customer’s discretionary goodwill, with the historical consequence, familiar from the restaurant trades, that the buyer’s obligation becomes variable with the customer’s mood while the headline rate quietly adjusts downward in expectation of the supplement. None of this is incidental colour. It is the daily machinery by which the value of labour power in the sector is set, contested, and disguised, and a reader who can see incentives as wage policy has acquired one of the transferable instruments this article exists to hand over.

Is the gig economy new? Day labour, outwork and the putting-out system

Novelty is the gig economy’s most successful marketing claim, repeated by celebrants and critics alike, and it is mostly false. Casual hiring by the day and the hour is older than the factory; the dock labourer shaping up at the gate each morning, uncertain of work by noon, was a stock figure of nineteenth century social investigation, and the long campaign for the decasualisation of dock work was a campaign against precisely the insecurity that platform rhetoric redescribes as flexibility. Outwork, the performance of industrial tasks in the home by workers owning simple tools and paid by the piece, clothed and fed whole districts in the garment and boot trades, and was analysed in its own time as a system for shifting costs, evading factory law, and depressing wages below the factory standard. The putting-out system, in which a merchant supplied materials and collected product from dispersed producers, was the dominant form of manufacture for centuries before the factory displaced it, and its producers owned their looms and workshops throughout, without that ownership preventing historians or contemporaries from tracing their subordination to the merchant’s control of materials and markets. Even the rating has ancestors, in the character references without which a domestic servant could not move between households, documents owned in effect by the previous employer and decisive for the next engagement. What the platform adds to these ancestors is real but specific: the automation of allocation and discipline, the scale and speed of the matching, the extraction of data from every transaction as a by product that becomes an asset in its own right, and the legal engineering that presents the whole as a software service. Those additions change the intensity, the reach, and the measurability of casual and outworked labour. They do not change its genus, and a class analysis gains authority by placing the phenomenon in its lineage, since a form that has recurred under every technological dispensation is unlikely to be explained by the technology alone.

The lineage also disciplines the analysis’s predictions. Outwork and casual hiring were not transitional embarrassments on the way to universal factory employment; they persisted wherever they served capital’s purposes, expanded in some periods, contracted in others, and coexisted with the standard employment relation throughout its supposed golden age. The secure, full time, benefited job that platform work is said to be displacing was itself, as the series discussion of precarity and the working class argues, a historically brief and geographically narrow settlement rather than the natural state of labour under capitalism. Platform work is therefore best analysed not as the future arriving early but as one pole of a permanent repertoire, the casual pole, refurbished with better instruments. That placement matters for the class question, because it removes the temptation to invent a new class for what is, in lineage and structure, the oldest form of the working class there is: labour hired by the task, at the gate, by the piece, from workers who own their tools and nothing else that matters.

The honest scale point: how much work is this?

Credibility requires the article to say plainly what advocacy on both sides obscures: platform work, measured as a share of total employment, is small in most economies, and measured as a share of working hours and earnings it is smaller still, because a large fraction of platform workers participate briefly, intermittently, or as a supplement to other income. Careful counts, whether built from tax records, labour force surveys with platform specific questions, or the platforms’ own disclosures, disagree with one another by wide margins, chiefly because they count different things: anyone who earned anything through a platform in a year is a far larger population than those for whom platform work is the main job in a given week, and both figures circulate in public argument, usually without the distinction attached, each selected for the conclusion it supports. The analyst’s obligation is to refuse the selection game. What can be said without a number is that the main-job platform workforce is a small minority of employment in every careful count, that participation in the looser sense is considerably wider, and that the sector’s significance is out of proportion to both because it is a laboratory: the contractual forms, the algorithmic techniques, and the cost shifting structures pioneered under the platform label migrate into conventional sectors, into warehousing, retail scheduling, care provision, and professional contracting, where their cumulative reach is far larger than the platform sector itself. A class analysis of the gig economy is therefore not a claim that the economy has become gig. It is a claim that the gig form makes visible, in concentrated and inspectable shape, mechanisms that operate more diffusely elsewhere, and that the classification worked out here transfers to those settings.

Why do estimates of the gig economy’s size disagree so widely?

Estimates disagree because they count different populations under one label. Counts of everyone who earned anything through a platform across a year produce large figures, while counts of people whose main job in a given week is platform work produce small ones, and surveys, tax data, and company disclosures each capture a different slice of the same loose phenomenon.

The scale point has a second edge, against the analysis’s own enthusiasts. A small and often transient workforce, much of it combining platform tasks with jobs, study, or benefits, is not the classical proletariat of the textbooks in its sociological texture, whatever its class position, and rhetoric that addresses it as the vanguard of anything should embarrass its authors. Class position is a structural location, not a measure of numbers, cohesion, or historical destiny, and the analysis in this article claims the first and nothing about the rest. The tradition has made its worst errors by sliding from position to destiny without evidence, and the discipline of the scale point, small share, honestly stated, is a guard against repeating the slide on this terrain.

The reserve at the gate: oversupply as a condition of the system

A feature of platform labour markets that participants report and the analysis should explain is the persistent sense of oversupply: too many drivers for the journeys available at the posted rates, too many couriers circling the restaurant districts, earnings per hour sagging whenever recruitment has run ahead of demand. Oversupply is sometimes treated as a platform error, a failure to balance the market the platform claims to optimise. The class analysis suggests the opposite reading. A standing surplus of available workers performs functions for the gateway’s owner that a balanced market would not. It guarantees the customer the instant availability on which the service’s appeal rests, since a pool of idle workers distributed across the city is the physical form of the three minute arrival time. It disciplines acceptance, because the worker who declines a poor task knows the allocator has alternatives and that the algorithm’s memory may not be neutral about refusal. And it holds the per task rate at the level the most pressed segment of the pool will accept, since the rate need only recruit the marginal worker, not satisfy the average one. The reserve army analysis developed elsewhere in this series, in its treatment of the relative surplus population, described exactly such a mechanism at the scale of the whole economy: a pool of labour available on call, whose existence regulates wages in the occupied sector and whose members bear the cost of their own availability. Platform recruitment automates the pool’s maintenance. Onboarding is continuous and nearly costless to the platform, the pool’s carrying costs, the vehicles, the insurance, the idle hours, are borne by the pool’s members, and the boundary between the employed and the reserve is redrawn at the level of the hour rather than the job, so that a single worker oscillates between the two conditions several times in a day. To be logged on and unallocated is to be in the reserve; to be dispatched is to leave it; and the wage for the dispatched intervals must, over the week, reconcile the worker to the unpaid intervals as well, a reconciliation the per task form is designed to obscure. The gig economy did not invent the industrial reserve. It miniaturised it, put it in the worker’s pocket, and arranged for the worker to pay its upkeep.

Who does the work: composition, migration and gender

Class position is a location, but the people who occupy a location arrive there by routes that shape the politics of the position, and the platform workforce’s composition deserves the same care as its classification. Recruitment into the dispatch sectors draws heavily on workers for whom conventional employment is obstructed: migrants whose qualifications are unrecognised or whose status makes formal hiring slow, workers whose caring obligations break standard rotas, students, the semi retired, and those rationed out of primary employment by discrimination, disability, or criminal records that gatekeepers screen for and platforms, in their growth phases, screened less. The low barrier at the gate is therefore double edged in a way the analysis must hold steadily. It is a genuine opening for workers the conventional labour market treats harshly, and the testimony that values it is often testimony about the alternatives those workers actually faced, not about platform work in the abstract. It is simultaneously the mechanism by which the platform assembles a workforce with weak outside options, whose reservation rate is low and whose capacity to refuse is correspondingly thin. Dependence is most complete where the gateway is also the only door, and the sector’s composition means that for a significant share of its workers, it is. The care and domestic rows add the dimension of gender to the same structure: platform cleaning and care recruit from the same heavily female, heavily migrant workforce that staffed domestic service through every previous dispensation, and the platform form modernises the recruitment, vetting, and payment of that workforce while leaving its social valuation, the low price attached to work coded as unskilled because it resembles unpaid household labour, substantially as it found it. An analysis that classified positions without noting who is sorted into them, and by what pressures, would be formally correct and substantially empty, because class in the living sense is always position plus composition, and the composition explains both the sector’s resilience and the specific bitterness of its disputes, in which the demand for recognition, of the work as work and of the worker as a worker, is never far below the demand for money.

Platform income: profit, rent, or a share of surplus value?

The platform’s own income poses the complementary classification problem, and the debate about it sharpens the class analysis of the workers. Three accounts circulate. On the first, the platform’s take is profit in the ordinary Marxist sense, a share of surplus value: the workers produce value in the service, the customer pays a price exceeding the wage equivalent the workers receive, and the platform appropriates the difference through its position in the circuit. On the second account, the platform’s take is rent. The platform owns a scarce gateway, the network of users and the data that makes matching possible, and charges for access to it as a landowner charges for access to fertile soil or a well placed site; on this reading the take is a deduction from value produced elsewhere, and the platform’s position resembles the monopoly of a necessary condition rather than the organisation of production. On the third account, the take is a merchant’s margin, the earnings of commercial capital, which in the classical scheme performs the necessary labour of circulation, buying and selling, matching and settling, without producing value, and is paid its share of surplus value for performing it. The accounts are not interchangeable labels. Each implies a different answer to where the platform’s income originates, and therefore a different analysis of what would change if the platform’s share were compressed: redistribution of a fixed surplus, abolition of a toll, or payment for a service that someone must perform under any system.

The fuller treatment of the platform case, including the dispute about whether user activity itself produces value, is given in the series article on platform capitalism from a Marxist standpoint, and the general form of the argument is the subject of the rent versus profit debate. For the purposes of class analysis, the decisive observation is that the three accounts converge on the worker’s position while diverging on the platform’s income. Whether the take is profit, rent, or commercial margin, it is extracted at the point where the worker meets the market, it is proportional to the volume of the worker’s labour mediated, and it depends on the platform’s ownership of the access the worker lacks. Even the rent reading, which is the most favourable to the platform’s self image as a mere venue, casts the worker as the party whose activity generates the flow from which the toll is taken, and a landowner’s tenant farmer, the classical figure for rent relations, owned more of the decisive means and set more of the terms than a ride hailing driver does. The income debate matters for many questions, for how platform wealth should be theorised, taxed, and contested, but it is a second order question for classification, and noting that protects the analysis from a common derailment, in which a genuine and interesting dispute about the platform’s revenue is used to postpone indefinitely the answerable question of the worker’s position.

The position it is confused with: the petty bourgeoisie

Because the ownership puzzle turns on the boundary between the working class and the petty bourgeoisie, the neighbouring position needs its own portrait, drawn from the series treatment of the petty bourgeoisie and the middle class. The petty bourgeois producer, in the classical scheme, owns means of production and works with them personally, employing few or no others: the independent artisan, the small shopkeeper, the family farmer, the self employed tradesperson with a personal clientele. The position is defined by a conjunction that is easy to state and was once easy to find: ownership of the productive assets the work requires, control over the labour process those assets serve, direct access to the market where the product is sold, and appropriation of the whole proceeds subject to the costs of materials and overheads. From the conjunction flow the position’s classical characteristics: the proprietor’s fierce defence of independence, which is real because the independence is real; the tendency to work longer hours than any employer could impose, self exploitation in the literal sense; the vulnerability to larger capitals that squeeze the small producer through prices, rents, and credit while leaving formal independence untouched; and the political volatility of a class that experiences the system simultaneously as its livelihood and as the force crushing it. The portrait matters here because it gives the classification something to be wrong about. If platform workers were petty bourgeois, they would exhibit the conjunction and its characteristics. They exhibit instead a systematic subtraction from the conjunction. The assets are owned, though financed and cost shifted. The control over the labour process is partial at best, ceded to the allocation and pricing system wherever the platform’s model requires it. The access to the market is not direct but rented by the transaction, and terminable without appeal. The proceeds are not appropriated whole but shared in advance with the gateway’s owner at a rate the owner sets. What remains of the petty bourgeois form is the solitude, the personal equipment, and the legal label, the costume of the position without its substance, and the characteristics follow the substance rather than the costume: platform workers in dispute demand what workers demand, higher rates per task, paid waiting time, protection against arbitrary deactivation, rather than what proprietors demand, lower fees in the abstract, access to more markets, freedom from regulation of their trade.

The subtraction analysis also locates the genuine exceptions with precision. Where a platform participant does hold the conjunction, the marketplace seller with a brand and a following, the skilled freelancer with a waiting list, the tradesperson whose platform is a diary service subordinate to a direct practice, the classification is petty bourgeois and the analysis should not flinch from it, even when the income is smaller than a driver’s. Class is not a compliment distributed to the deserving. It is a map of positions, and a map that refuses to mark real proprietors because their work arrives through an application would be as worthless as a map that marks every driver a proprietor because the car has a registration document. The boundary runs through the platform economy, not around it, and the grid’s mixed rows are the map drawn honestly.

The strongest objection: discretion, preference, and the charge of condescension

Every preceding section now faces the objection that the brief for this article requires to be stated at full strength, and it is a strong one, with an empirical and a moral strand. The empirical strand begins from the discretion platform workers undeniably exercise. No foreman assigns the driver’s shifts. The courier chooses which days to ride, the cleaner which bookings to accept, the freelancer which commissions to refuse, and large numbers use exactly that discretion as the arrangement’s chief virtue, fitting platform work around study, around children, around another job, around a fluctuating capacity to work that no rota would accommodate. Multi-apping, running several platforms at once and taking work from whichever offers best at the moment, is discretion extended into a rudimentary form of market making on the worker’s own behalf. Surveys of platform workers regularly find substantial minorities, and in some sectors majorities, reporting satisfaction with the arrangement and a preference for it over the available conventional jobs, and ethnographic work finds the preference intelligently defended by the people who hold it, in terms of autonomy from supervisors, escape from workplace politics, and control over the week’s shape. The moral strand then presses the point home. A framework that classifies these workers as proletarians against their own understanding of their situation, that treats their stated preference as a veil over their real position, repeats the condescension problem that haunts the class consciousness debate: the analyst claims to know the worker’s interests better than the worker does, and the claim is unfalsifiable in the analyst’s hands, since every disagreement becomes further evidence of the veil. Worse, the classification is not innocent. It arrives with a politics attached, implying organisation, regulation, and employment status that many of the classified do not want and some actively resist, fearing, with reason, that the flexibility they prize would be the first casualty of the remedies proposed in their name. On this objection, the decisive-means test is a scholastic device for overriding the testimony of the people it describes, and the honest analyst would start from that testimony instead.

The objection must also be given its strongest theoretical form, which concerns the criterion itself. Discretion over when to work is not a detail of the platform arrangement; it touches the definition of the wage relation, in which the capitalist buys labour power for a determinate period and directs it throughout that period. A seller who decides whether to sell at all on a given day, who sells by the task and withdraws at will, looks less like the classical wage labourer than like the independent commodity producer the theory distinguishes from the wage labourer, and several serious writers within and beyond the Marxist tradition have argued on exactly this ground that platform work revives commodity production by individuals, a market of small sellers coordinated by a platform that functions as their collective shopfront, closer to a bazaar with software than to a factory without walls. On that reading the platform is a new kind of market institution and its participants are a new kind of petty commodity producer, and the class analysis offered in this article is the misapplication of an industrial framework to a post industrial form it cannot see. That is the objection at full strength: empirically grounded in real discretion and real preference, morally grounded in respect for testimony, theoretically grounded in the wage relation’s own definition.

The reply: position is not preference, and the concession that stands

The reply begins with a distinction that dissolves the moral strand without dismissing it. Class analysis classifies positions, not experiences, and a position is defined by relations that hold whether or not they are felt: dependence on a gateway for access to work, subjection to prices set elsewhere, exposure to exclusion without recourse. A tenant farmer may prefer tenancy to the landless labourer’s lot and may say so sincerely; neither the preference nor the sincerity alters the tenure. A worker’s evaluation of an arrangement answers the question of how the arrangement compares with the alternatives actually available to that worker, which is a real and important question, and a different one from where the arrangement stands in the relations of production. Satisfaction, in particular, is measured against the local baseline, against the warehouse shift, the call centre rota, the care job with its unpaid travel time, and a verdict of better than those is fully compatible with the position being what the analysis says it is. Nor is the testimony as one sided as the objection’s presentation suggests: the same survey literature that records satisfaction with flexibility records dissatisfaction with pay levels, unpaid waiting, opaque allocation, and deactivation without appeal, and the same workforce that defends the arrangement’s freedoms organises, as the delivery section noted, against its rates. Workers are perfectly capable of valuing an arrangement’s discretion and disputing its terms simultaneously, because they are evaluating different features, and an analysis that listens to the whole testimony rather than selecting the convenient half hears no contradiction.

The theoretical strand requires a concession, and the concession should be made without throat clearing. Discretion over scheduling is a genuine difference from classical wage labour, not an illusion, and it marks a real variation in how labour power is engaged: by the task and the episode rather than by the week and the year. The analysis claims that the variation is a variation, in the duration and granularity of the engagement, not a difference in the relation engaged in, and the ground for the claim is the direction of dependence during every episode however brief. When the driver is working, the price is set, the allocation is governed, the standard is enforced, and the customer belongs to the platform; between episodes the driver is free, as the day labourer at the dock gate was free between mornings, as the outworker was free between bundles, as every casual worker in the history of the wage relation has been free in the intervals of not being hired. Casualness is not the opposite of the wage relation. It is one of its oldest temperatures, and the freedom to decline work one cannot afford to decline is a freedom the tradition has always recognised as compatible with dependence, because dependence is measured by the consequences of refusal sustained over time, not by the formal existence of a choice at each moment. The commodity producer reading fails, finally, on the point where it can be tested: the classical petty producer sells a product whose price the market makes from many sellers’ independent offers, while the platform worker faces a price the platform computes and varies unilaterally, with no offer of the worker’s own ever reaching the customer. A bazaar stallholder names a price. A driver is named one. That asymmetry, small enough to state in a sentence, is the difference between a market of producers and a labour market wearing a market’s clothes.

What survives of the objection, after the reply, is its warning rather than its conclusion. The warning is that classification carries no automatic politics, that a worker’s position does not dictate the worker’s programme, and that remedies designed from the classification alone, indifferent to the discretion workers prize, will be resisted by their intended beneficiaries and will deserve to be. The analysis accepts the warning and builds it into the closing assessment: what follows from a class position is a set of questions about power and dependence, not a prefabricated set of demands, and the terms on which platform work should be reorganised are a matter for the workers in it, whose testimony about flexibility is evidence about what any defensible settlement must preserve. Condescension consists in overriding testimony about experience. It does not consist in maintaining, against a contractual label, an analysis of structure that the testimony, taken whole, tends to confirm.

Multi-apping and the portfolio worker

The strongest empirical refinement of the objection comes from a practice the reply must absorb rather than brush past: multi-apping. Experienced platform workers commonly hold accounts with several platforms and work them in combination, watching the rival offers, logging off the platform whose rates sag and onto the platform whose surge rises, treating the applications as competing buyers in a spot market of their own construction. The practice is real arbitrage, and it changes the worker’s day in ways the single platform portrait misses. A worker who can credibly threaten, by conduct rather than by words, to take the next hour elsewhere possesses a margin of bargaining power that the captive single platform worker lacks, and platforms know it, which is why exclusivity incentives, bonuses conditional on volume with one service, and gentle penalties for divided attention recur in the incentive architecture. Multi-apping also supplies the objection’s best anecdotal material, the worker as tactician, reading four screens like a trader, contemptuous of the suggestion that the relation is one of dependence.

The analysis absorbs the practice by locating it correctly. Arbitrage between buyers presupposes buyers whose terms the arbitrageur does not set. The multi-apping driver chooses which of several unilaterally posted prices to accept at each moment and sets none of them; the freedom is the freedom of the seller in a market of price makers, which improves the seller’s outcome at the margin while leaving the structure of price making untouched. The dock labourer who presented at several gates, the outworker who carried bundles to two merchants, exercised the same margin, and it coexisted with dependence because dependence is a property of the worker’s relation to the market as a whole, not of loyalty to one gate within it. What multi-apping genuinely changes is the gradient: the skilled arbitrageur with several accounts, high ratings on each, and the local knowledge to play the systems against one another stands measurably further from the pole of total dependence than the single platform novice, and a fine grained class analysis should register position as a location on such a gradient rather than as a cell in a table. Registration of the gradient is not retreat from the classification. The gradient runs within the working class’s territory, between its most and least dependent forms, and it reaches the boundary only where the worker’s alternatives cease to be other gateways and become a market of the worker’s own, which is the portability condition the grid’s freelancing row describes.

The competing accounts: entrepreneurship, flexibility economics, and the labour law reading

Three non-Marxist accounts of the phenomenon compete with the analysis offered here, and the template for application articles requires the strongest of them to be met rather than caricatured. The entrepreneurship account, which is the platforms’ public philosophy and also a sincere belief of many participants, holds that the platform has lowered the barriers to running a business until the business and the individual coincide: the driver is a transport company of one, the courier a logistics firm with a bicycle, and the platform a supplier of services, customers, payment processing, brand trust, to these micro enterprises. Its strength is its fit with the genuine discretion and the genuine asset ownership; its weakness is the conjunction test of the previous section, since a business that cannot set its prices, cannot keep its customers, cannot sell itself, and cannot refuse its sole supplier’s terms without ceasing to trade is a business in a sense that empties the word. The flexibility economics account, the mainstream labour economics reading, treats platform work as a matching improvement: digital allocation reduces the friction of casual labour markets, workers reveal by participation that the package beats their alternatives, and the policy questions are about portability of benefits and the tax treatment of the income rather than about class. This account is the most formidable of the three because its evidence, revealed preference and measured earnings against local alternatives, is real, and the reply to it is not denial but reframing: it describes the transaction’s surface accurately while declining to ask who owns the matching infrastructure and what that ownership is worth, which is the question the class analysis exists to ask. The labour law reading, finally, treats the phenomenon as a classification problem to be solved by better statutory tests and enforcement, and within its own terms it is indispensable, since rights depend on it; its limitation is the one identified in the third answer, that legal categories are revisable instruments which track the underlying relation only loosely and sometimes by accident. The Marxist account does not refute these three so much as locate them, as descriptions of the phenomenon’s appearance, its market mechanics, and its legal shadow, each partial, and each compatible with the structural claim they are usually deployed to avoid.

What the framework reveals, and where it strains

What the framework reveals, gathered from the preceding sections, can be stated compactly. It reveals that the gig economy’s central drama is not technological but proprietary: the ownership of a new kind of productive asset, the coordinated network of demand, data, and dispatch, and the use of that ownership to reorganise old forms of casual and piece paid labour at unprecedented scale and measurability. It reveals the wage relation operating under contractual forms designed to make it invisible, and it gives the reader instruments, the decisive-means test, the subsumption inversion, the piece wage analysis, that recover the relation from the forms without denying anything the forms genuinely contain, the discretion, the assets, the episodic freedom. It reveals the continuity between the platform and the putting-out merchant, the frame renting master, and the dock gate, and continuity is itself a revelation in a debate addicted to novelty. And it reveals why the politics of the sector take the shapes they do, why disputes ignite over fees, waiting time, and deactivation rather than over the platform’s existence, because those are the wage relation’s pressure points under its platform form.

Where the framework strains should be stated with equal compactness, because the strain points are where honest readers will test it. The extraction mechanism is clearest where a service is produced and sold in one motion, as in transport, and murkier where the platform’s revenue depends on advertising, data, or the subsidised conquest of a market, phases in which a platform may mediate enormous labour while appropriating little from it directly, sustained by investors purchasing future position. The analysis handles those phases awkwardly, and the rent account of platform income does some of the work the surplus value account cannot. The framework also strains at the sector’s upper edge, where scarcity of skill reverses dependence and the platform becomes a vendor to producers rather than their gateway, and it strains, as every structural analysis does, to explain variation in experience: two drivers on the same system inhabit the same position and different lives, one treating the work as a bridge, one as a trap, and the difference lies in biographies, households, and local alternatives that class position frames but does not determine. Finally, the framework’s collective implications strain against the sector’s atomisation, which is the subject of the next section but one. None of these strains overturns the classification. Each marks a place where the classification is the beginning of the analysis rather than its end, which is what a framework is for.

Classification by courts: what law decides and what it cannot

The legal classification landscape deserves a consolidated description, because readers meet the topic through it and because the article’s discipline, describe the pattern and the tests, never a frozen outcome, is easiest to honour in one place. Across jurisdictions, the disputes share a grammar. Claimants argue that the reality of the relationship matches the statutory idea of employment or of the intermediate worker category where one exists: control by the platform over price, standards, and allocation; personal service expected in practice; integration of the worker’s activity into the platform’s business, since the platform sells journeys and deliveries, not software licences, to its customers. Platforms argue the contractual architecture: the freedom to log on and off, the absence of exclusivity, the worker’s equipment, the theoretical possibility of substitution, the multiplicity of platforms a worker may serve. Adjudicators applying functional tests have in a number of prominent disputes found for the claimants, reasoning that written terms must yield to the reality they purport to describe, and that a business which fixes the price and disciplines the performer has taken the incidents of employment while declining its obligations. Legislatures have then pulled in both directions: some writing the adjudicators’ conclusions into presumptions of employment for defined platform sectors, some creating bespoke intermediate statuses with partial rights, and some reversing adjudication outright by writing contractor status for application based work into statute, in one prominent case through a ballot initiative funded at a scale that itself became part of the argument about power in the sector. The direction of travel is therefore not single, and any sentence that reports the legal position of platform work in general is false somewhere. What can be reported is the persistence of the contest and its recurring subject matter, which is control, and control is the very variable the class analysis isolates on independent grounds.

For the analyst, the legal record’s deepest lesson is about the limits of formal categories under strategic pressure. Each legal test, once articulated, becomes a design specification: contracts are redrafted to gesture at substitution, allocation systems are adjusted to display choice, terms are rewritten to perform independence while the operational system retains the controls that matter. The cat and mouse sequence is not a scandal particular to platforms; it is the normal history of employment law at every boundary, from the homeworking outworkers whom factory acts struggled to reach to the agency arrangements of the warehousing sector. But it demonstrates why a class analysis cannot be subcontracted to the law. The sociological question, where the worker stands in the relations of production, is stable across the redrafting in a way the legal answer is not, because the underlying configuration, gateway ownership, price setting, disciplined allocation, changes only when the business model changes, not when its paperwork does. Readers who follow the litigation should therefore read judgments as evidence about the configuration, richest where courts describe operations in detail, and should resist the headline inference, in either direction, that a ruling has settled what platform work is.

Deactivation: dismissal in a relation that denies it has employees

Every employment relation reveals its structure at the moment of its ending, and platform work’s ending is deactivation, the closure of the account by the platform, frequently automated in its initiation and opaque in its reasoning, against which the contractual form provides no hearing, no notice period, and no severance, since formally there was nothing to sever. The grounds range across fraud flags generated by the system itself, customer complaints aggregated past a threshold, document expiries, background check renewals, and patterns the platform declines to specify, and the worker’s experience of the event has the distinctive character that runs through the ethnographic record: the discovery is made by opening the application, the explanation arrives as a category rather than a reason, and the appeal is addressed to the same system that decided. Whatever the legal label of the relationship, deactivation functions as dismissal, and its stakes are dismissal’s stakes raised higher, because the deactivated worker loses not one job among possible similar jobs but access to the market the account mediated, including the rating archive in which a working history was stored. Campaigns for deactivation due process, human review, stated reasons, compensation for wrongful closure, have become a standard item of platform worker organisation across jurisdictions, and their prominence is diagnostic. Workers do not campaign for due process against a supplier. A business whose customer delists its products has commercial remedies and other customers; the demand for a hearing is the demand of a party who understands the relation as one of continuing dependence terminable at the stronger side’s discretion, which is to say, as employment in everything the law declines to name.

Deactivation also completes the subsumption argument from an unexpected direction. A system that can dismiss without a manager, by threshold and flag, has automated the final function of supervision, and the automation matters analytically because it shows the disciplinary apparatus to be a property of the platform’s infrastructure rather than a set of human decisions the platform merely transmits. The foreman has not been abolished. The foreman has been generalised, distributed into allocation, rating, incentive, and exclusion, and rendered continuous and tireless in the process. When the analysis claims that algorithmic management changes the measurability and therefore the intensity of the labour process, deactivation is the extreme case of the same principle: even the sanction that ends the relation is measured, triggered, and executed inside the informational system that owns the decisive means.

Class consciousness among workers who never meet

The classical theory connected class position to class consciousness through the workplace: concentration in factories taught workers their common situation and gave them the daily association from which organisation grows. Platform work inverts the conditions. The workforce is dispersed by design, meets customers rather than colleagues, communicates with the platform through an interface that individuates every message, and experiences grievances, the changed formula, the vanished bonus, the sudden deactivation, as personal events before they can be recognised as common ones. The rating system adds a finer atomisation, setting each worker’s record against an invisible distribution of others, so that competition is continuous and its object is the worker’s standing with the allocator rather than any shared standard for the job. A theory that predicted consciousness from position mechanically would predict none here, and would be refuted by the record, because organisation has repeatedly emerged: courier collectives coordinating stoppages through messaging applications, driver associations contesting fee changes city by city, legal challenges funded and carried by groups of workers who first found one another online, and mutual aid structures for the deactivated and the injured that reproduce, in digital form, the friendly societies of an earlier casual workforce. The interesting question is not whether consciousness appears but what it fastens on, and the record suggests it fastens on the wage relation’s platform pressure points, the rate per task, the unpaid interval, the arbitrary exclusion, rather than on the platform’s existence or on ownership in the abstract. Organisation among gig workers therefore looks less like the factory movement’s demand for the means of production and more like the casual trades’ historical demands, a standard rate, a regulated gate, an end to arbitrary refusal of work, which is consistent with the lineage this article has drawn and a useful corrective to accounts that find the sector’s politics puzzlingly modest. The modesty is the politics of casual labour, and it has a long pedigree.

For the class analysis, the consciousness question closes the loop opened by the objection section. Workers’ self understanding is evidence, and the evidence here is double: platform workers defend their discretion in the language of independence and contest their rates in the language of labour, often in the same interview, sometimes in the same sentence. The analysis that classifies them as working class is not overriding that testimony but organising it, attributing the independence language to the real discretion the form contains and the labour language to the real dependence the form also contains, and locating both in a position whose structure generates exactly that divided consciousness. A position that produced a single, settled self description would be a simpler thing than any class position in history has ever been.

One working day, read through the analysis

Instruments are only as good as their handling, so consider a single illustrative working day, constructed from the common features of the dispatch sectors rather than drawn from any individual’s record, and read it through the categories this article has built. The driver begins the day as a member of the reserve, logged on at home, available and unpaid, the car’s loan payment and insurance running silently in the background as the day’s fixed deduction, the shifted constant capital asserting itself before a single journey exists. The first allocation arrives with a price attached and a destination withheld, and the acceptance is made in the few seconds the interface allows, under the knowledge that refusal is recorded; here is real subsumption at the level of the individual decision, the labour process specifying not only the task but the information on which the worker may act. The morning peak pays well and the surge map glows, and the driver works the incentivised hours the platform most needs covered, experiencing the alignment as personal cleverness in choosing a start time, while the pricing system experiences it as supply delivered to the minute; both descriptions are accurate, and their compatibility is the flexibility debate in miniature. Midday brings the trough the per task form renders invisible as a cost: forty minutes of circling and waiting between journeys, labour power held at the buyer’s disposition and paid nothing, the working day in the classical sense continuing while the paid day pauses. A passenger, late and anxious, rates the journey four stars for traffic the driver did not create, and the plebiscite of the customer inspector registers its small deduction from the asset the worker produces and does not own. The afternoon quest bonus assembles its quiet compulsion, each completed trip raising the cost of stopping, until the driver works two hours past the intended end of the shift to close a sequence the platform designed. The day closes with the accounts the form invites the worker to keep: gross receipts, prominently displayed by the application, against which fuel, depreciation, and the day’s unpaid intervals appear nowhere, so that the worker’s own bookkeeping, if it follows the application’s categories, reproduces the disguise the piece wage analysis describes. Every element of the analysis has appeared in the day without forcing: the decisive means in the allocation and the withheld destination, the shifted costs in the silent deductions, real subsumption in the managed information, the reserve in the unpaid availability, the rating in the traffic’s four stars, the piece form in the vanished midday. A framework that can narrate an ordinary day at that resolution, finding its categories in the hours rather than imposing them on the hours, has passed the test an application article sets itself.

What would change the classification: the test’s defeaters

An analysis that cannot say what would prove it wrong is a vocabulary, not an argument, so the conditions under which platform workers would cease to occupy a working class position should be stated as plainly as the classification itself. The decisive-means test supplies them directly: the classification changes when the decisive means move. Concretely, a platform arrangement would cross the boundary if workers could carry their ratings and customer relationships between platforms or into independent practice as of right, since the gateway would then cease to be a gateway and become a directory. It would cross if prices for the service were proposed by workers and accepted or declined by customers through the platform, with the platform confined to a transparent commission, since price setting is the daily signature of command over the labour process. It would cross if allocation were by open claim rather than by assignment, with full information, destination, distance, and pay, available before commitment, since the informational monopoly is the mechanism of real subsumption in the sector. And it would cross most completely under worker or municipal ownership of the dispatch infrastructure itself, the platform cooperative and the public platform experiments that recur in policy debate, in which the coordinating asset is held by the producers who use it and the surplus from coordination is theirs to distribute.

These defeaters do analytical work in both directions. Against the analysis’s critics, they show that the classification is an empirical claim about an arrangement, falsifiable by changing the arrangement, rather than a prejudice about a technology; nothing in the argument requires dispatch software to be anyone’s property in particular, and the same software under cooperative ownership would support a different classification of the same drivers doing the same journeys. Against the analysis’s careless friends, the defeaters expose how little of the structure is touched by the reforms most often demanded in the classification’s name. A higher minimum per task, welcome as it may be, leaves gateway, pricing, allocation, and rating exactly where they stand; it raises the wage within the relation without altering the relation, which is the oldest distinction in the tradition between movements on the surface of the wage and changes in the position of the class. The cooperative experiments themselves exhibit the countervailing difficulty honestly: the decisive means are decisive partly through scale, a dispatch network’s value grows with its pool of drivers and customers, and producer owned platforms must solve the capital and coordination problems of reaching that scale against investor funded rivals prepared to subsidise both sides of the market for years. That the difficulty is real does not rescue the investor owned form from the analysis. It explains the form’s dominance, which is a different thing, and explanation of dominance is what class analysis has always offered in place of the consolation that the dominant arrangement is also the natural one.

Partial movements along the defeaters’ direction are equally informative, because the test predicts their effects in advance. A portability rule for ratings, where regulators have entertained one, should measurably improve the exit position of experienced workers without disturbing the platform’s command of price and allocation, and the analysis therefore predicts a workforce somewhat less dependent and identically classified, a movement along the gradient of the multi-apping discussion rather than across the boundary. A transparency rule for allocation information should soften real subsumption at its informational edge while leaving the subsumption intact, since the system would still assign, price, and judge. Watching which reforms move which variable, and whether the predicted magnitudes appear, is how a classification earns the status of an analysis over time: by surviving contact with interventions it did not design, in a sector that redesigns itself continuously, and continuing to sort the redesigned forms correctly.

For students and exam candidates

This article supplies the strongest contemporary example available for any examination question on class, and it should be used as an example with an argument attached rather than as an illustration. The distinction that earns marks is the decisive-means test: ownership of the means of production that govern the labour process, not ownership of any asset, settles the classification, and the vehicle example wins the point in two sentences because the car is replaceable and the gateway is not. The standard trap is to argue from legal status, citing tribunal outcomes as though a statutory category were a sociological one; markers reward candidates who separate the questions and then note the corroboration. A second trap is the new class claim, which dissolves under the lineage evidence of outwork, piece payment, and casual hiring. Candidates preparing the topic should keep the grid, the three answers, and the subsumption inversion in a single organised set of notes, and readers who save the grid and build a citation-linked set of class notes on VaultBook will find the example ready to deploy when the question pairs class with technology, work, or inequality.

For teachers and seminar leaders

The misconception that dominates a class on this topic is that ownership is a matter of title to objects, so that the owned car settles the argument before it starts. The question that surfaces it is disarmingly concrete: ask the seminar which the driver could more easily replace within a week, the car or the account, and let the asymmetry do the teaching. The example that resolves the residual confusion is the paired plumber, identical in tools and van, independent with a personal clientele in one version and platform dependent in the other, since the pair isolates the customer relationship as the decisive variable with everything else held constant. A productive seminar exercise assigns the grid’s rows to groups and requires each group to defend a classification for its row using all three answers, then to state the strongest objection to its own verdict; the freelancing row reliably produces the best argument, because it is the row where the evidence genuinely divides. Teachers should also expect the flexibility objection to arrive with personal testimony behind it, from students who do this work, and the seminar is richer when that testimony is treated as evidence about experience to be organised rather than as an error to be corrected.

For researchers, historians, and journalists

The journalist and policy layer carries the heaviest verification burden in this article, because the vocabulary of the gig economy is contested in print daily and the figures are weapons before they are measurements. Verify the population before repeating any number: platform companies report registered accounts, active users under definitions they choose, and participation totals that count a single delivery in a year, while labour force instruments count main jobs in a reference week, and a paragraph that moves between those populations without marking the move misleads however accurate each figure is in isolation. Verify the legal position in the specific jurisdiction and at the specific date of writing, since classification in this sector is a moving target of judgments, presumptions, bespoke statutes, and ballot reversals, and a ruling described without its jurisdiction is a rumour with a citation. Verify, above all, the operational facts before characterising the relation: whether destinations are shown before acceptance, whether refusal carries allocation consequences, whether rates are negotiable in any direction, whether accounts can be sold or bequeathed, because the class question turns on those mechanics and they differ between platforms and change between versions of the same platform. Researchers will find the primary record unusually rich and unusually treacherous: tribunal findings and disclosed internal documents give the most reliable accounts of the labour process, platform published research gives scale without independence, worker testimony gives experience with selection effects, and the tradition’s categories, tested here against the hardest case, give the questions that organise the rest. Historians should note the recurrence of the form, since the putting-out and casual hiring archives are the control group against which claims of novelty can be measured rather than merely asserted.

What the analysis establishes, and what remains open

The analysis establishes a classification and a method. The classification: platform workers of the dispatch type, ride hailing drivers, food delivery couriers, microtask workers, and platform dependent cleaners and carers, occupy a working class position, because the means of production decisive for their labour processes, the matching infrastructure, the data, the pricing power, and the customer relationship, are owned by the platform, while the assets the workers own function chiefly as shifted costs of production. The method: the decisive-means test, which asks whose withdrawal ends the worker’s access to production, which assets can be replaced and which cannot, and who sets the price and the standard, and which therefore classifies the mixed cases, the marketplace proprietor, the scarce skilled freelancer, with the same instruments rather than by exception. Around the classification, the article has established the subsumption inversion, that platform capital really subsumes a labour process it declines to equip, transforming work through information while shedding the costs of the instruments; the piece wage reading, that platform payment is the classical piece form perfected by software and presupposes the wage relation it disguises; and the legal corroboration, that functional legal tests converge on the same configuration for their own purposes without settling the sociological question. What remains open is considerable: the correct theorisation of platform income between profit, rent, and commercial margin; the trajectory of legal classification as legislatures continue to intervene; the forms organisation will take in a workforce designed for dispersion; and the migration of platform techniques into the conventional economy, where the classification worked out here will be tested against workplaces that still look like workplaces. The analysis closes where an application should, with the phenomenon classified, the instruments named, and the open questions stated as questions rather than concealed as answers. For the series as a whole, the article’s place is at the point where the class cluster meets the present: the criterion defended in the cluster’s theoretical articles is worth only what it can classify, and the gig economy was the hardest classification on offer, a workforce holding real assets under real direction, dispersed by design, divided in its own account of itself. That the criterion, sharpened into the decisive-means test, classifies the case without violence to the evidence, and classifies the neighbouring cases, the proprietor with a clientele, the professional with a waiting list, the cooperative that might be built, differently and for stated reasons, is the strongest available evidence that the criterion is an instrument rather than an inheritance. Instruments are judged by the work they do. This one has done its work here, and it is left in the reader’s hands in working order.

Frequently Asked Questions

Q: Are gig workers working class or self employed?

They are, in the main, both in different vocabularies: self employed as a contractual and legal description, working class as a position in the relations of production. The decisive-means test settles the second description. The platform owns the matching infrastructure, the data, the pricing, and the customer relationship, without which the worker cannot reach the market at all, while the worker owns equipment that functions chiefly as a cost of production. A minority of platform participants, those with portable reputations and direct clienteles, are genuinely independent producers, and the analysis classifies them differently rather than forcing one label on a divided sector.

Q: Who owns the means of production in platform work?

Ownership is split, and the split is the puzzle. The worker owns the physical instruments: the vehicle or bicycle, the phone, the tools, the materials. The platform owns the coordinating assets: the dispatch and matching system, the accumulated demand and performance data, the payment channel, the brand the customer trusts, and the rating archive that governs future access. Class analysis asks which set is decisive for the labour process, and the coordinating set is, because its withdrawal ends the work entirely while the physical instruments can be rented, borrowed, or replaced without changing the worker’s dependence on the gateway.

Q: Is algorithmic management a form of real subsumption?

Yes, in the dispatch type of platform work, and the claim is specific. Real subsumption means capital remaking the labour process itself rather than merely commanding its product. Algorithmic allocation decides which task the worker performs and in what order, pricing formulas set the value of each unit, routing and timing govern the path and pace, and rating systems enforce standards continuously. The work of driving or delivering as the platform organises it could not exist without that designed system, which is the mark of real subsumption achieved through information rather than through machinery the capitalist owns.

Q: Are platform workers a new class?

No, and the claim of novelty is the sector’s most successful marketing. The configuration, workers owning simple tools, paid by the piece, dependent on an intermediary who controls access to the market, describes the outworkers of the putting-out system, the domestic workers of the sweating trades, and the casual labourers of the dock gate, all of whom were classified within the existing scheme without a new category. What is new is the instrument, automated allocation and measurement, and the scale it permits. New instruments applied to an old relation produce a variation of the working class, not a successor to it.

Q: How have courts classified gig workers?

Inconsistently, and the inconsistency is itself the finding. Courts and tribunals applying functional tests, control, personal service, integration, the reality behind the written terms, have in many prominent disputes found worker or employee status, reasoning that a platform which sets the price and disciplines performance has taken the substance of employment. Legislatures have responded in opposite directions, some writing presumptions of employment for platform sectors, some creating intermediate statuses, and some enacting contractor status for application based work. The pattern to report is a continuing contest conducted through the variable of control, not a settled legal position.

Q: Does owning your own vehicle change your class position?

No, because class position follows the decisive means of production rather than any asset whatever. The vehicle is replaceable, rentable, and useless as a lever: it cannot set the price of a journey, cannot carry a customer relationship with it, and cannot be sold as a going concern together with the work it serves. Its ownership chiefly transfers the costs of capital equipment, purchase, fuel, insurance, depreciation, from the platform to the worker. The asset whose loss would end the worker’s access to the market is the platform account, and the account belongs to the platform.

Q: Is platform income profit or rent?

The question is genuinely disputed, and the answer matters less for classification than the dispute suggests. On one account the platform’s share is profit, a portion of surplus value produced in the service and appropriated through the platform’s position in the circuit. On another it is rent, a toll charged for access to a scarce gateway, the network and its data, deducted from value produced elsewhere. On a third it is a merchant’s margin for performing circulation. Each account assigns the worker the same position, since each describes a flow extracted at the point where the worker meets the market through an asset the worker does not own.

Q: Can gig workers develop class consciousness?

Yes, and the record shows forms of it, though shaped by the sector’s dispersion. Workers who never share a workplace have organised stoppages by coordinated log offs, formed associations to contest fee changes, funded legal challenges collectively, and built mutual aid for the deactivated. The consciousness that emerges fastens on the wage relation’s pressure points under platform conditions, the rate per task, the unpaid waiting interval, the arbitrary exclusion, rather than on ownership in the abstract. That pattern matches the historical politics of casual and outwork trades, whose demands were standard rates and regulated access long before they were anything larger.

Q: How big is the gig economy really?

Smaller than its share of argument, and any single figure should be distrusted. Counts differ because they measure different populations: everyone who earned anything through a platform across a year is a large population, while those whose main job in a given week is platform work are a small minority of total employment in every careful count. The sector’s importance is disproportionate to either figure because its contractual forms and management techniques migrate into conventional employment. An honest sentence gives the structure, main job platform work is a small share, participation in the loose sense is wider, and declines the false precision of a percentage.

Q: Is the gig economy a return to piece work?

Yes, in the exact sense Marx’s analysis of piece wages gives the term. Platform payment is by the unit, the trip, the drop, the task, and piece payment is a converted form of time payment: the rate is calibrated so that the average worker at the required intensity earns the equivalent of labour power’s value across the period, while the form disguises the working day, leaving waiting and travelling time unpaid and uncounted. Piece payment also enforces intensity without supervisors, since stopping means earning nothing, and invites an intermediary’s cut on every unit, which is precisely the platform’s share.

Q: How do employment tribunal rulings differ from a class analysis of the same work?

They answer different questions with similar evidence. A tribunal decides a statutory question, whether this worker falls within a legal category carrying specified rights, using tests drafted for administrability and revisable by any legislature that dislikes the outcome. A class analysis decides a structural question, where the worker stands in the relations of production, using criteria that do not change when a statute does. The two often converge because both examine control, dependence, and the reality behind the contract, and the legal record is valuable evidence for the analyst. But a ruling can be reversed by a ballot measure, and a class position cannot.

Q: What should a journalist verify before repeating a platform’s claim about the size of the gig economy?

Three things, at minimum. Verify the population: registered accounts, people active under the company’s chosen definition, and people whose main job is platform work are different numbers differing by orders of magnitude, and company claims use the largest available. Verify the period and the intensity threshold, since one delivery in a year and forty hours in a week both count as participation in loose definitions. Verify the source’s independence, because platform published figures are advocacy instruments, while tax records and labour force surveys, though imperfect, are not selling anything. A claim that survives all three checks can be reported with its definition attached.

Q: Why does owning tools matter less than owning the marketplace in platform work?

Because tools without a market are inert, and the marketplace is where access, price, and standards are decided. A courier who owns a bicycle outright still cannot convert it into income without orders, and the orders arrive only through the platform’s allocation, at the platform’s price, under the platform’s discipline. The marketplace owner can replace any individual tool owner from the pool of applicants without interrupting the service. Ownership of the marketplace therefore confers command over the labour process of everyone who works through it, while ownership of tools confers only the obligation to maintain them. Decisiveness, not price or size, is what class analysis counts.

Q: How does piece payment disguise a wage relation in gig work?

By renaming every element. The wage becomes a fee, the working day becomes a set of independent tasks, the employer’s direction becomes the application’s suggestions, and the unpaid portions of the day, waiting for allocation, travelling to the zone, queuing at the restaurant, become the worker’s private time rather than labour the buyer declines to pay for. The disguise works because the payment genuinely is per piece; what it conceals is that the piece rate is calibrated against time and intensity, so that the worker sells labour power by the hour at one remove, under a form that makes the sale look like a series of tiny businesses trading.

Q: What would class consciousness among couriers have to organise around?

Around the points where the wage relation is felt under platform conditions. The rate per drop and the formula that changes it without negotiation; the unpaid intervals, waiting at restaurants and between orders, that the per task form renders invisible; the transparency of allocation, including whether destinations and pay are known before acceptance; and protection against deactivation without a human hearing, since exclusion from the account is exclusion from the market. Those demands correspond to the classical programme of casual trades, a standard rate, paid time, and a regulated gate, and they are the demands courier organisation has in practice produced.