The ceiling is hours, and it arrives sooner than people expect

Every job in this trade needs you present for it, which means your income has a hard ceiling and the ceiling is made of hours. Most people find it in the second year and are surprised, because the arithmetic they did in month one was the wrong arithmetic.

This is the capacity view of the work: how many jobs fit in a day, how much of a week is not billable at all, and what happens when you reach the top of it.

The arithmetic everyone does first

Take your time per job, divide it into a working day, multiply by five, multiply by your fee. It produces a large and completely fictional number.

The fiction has three sources, and they compound rather than add.

The first is that time per job is not the time a job takes. It is the time the work takes, and the work is perhaps two thirds of the job.

The second is that a working day does not contain eight working hours in a one-person service business, and it has never contained eight in any business.

The third is that jobs do not queue up neatly to fill the day. They arrive when buyers send them, which is not when you are free.

Run the honest version and a figure that looked like a comfortable full-time income becomes something more like a modest one, at a level of effort that is not sustainable across a year.

The unbilled half of the week

The single largest correction is the time that produces no invoice.

Enquiries that do not convert. Messages before, during and after every job. Reading briefs and asking the one clarifying question. Delivery, handover, and the follow-up that follows the handover. Records, reconciliation, chasing a payout, reading a terms-change email. Rewriting a listing, building a worked example, answering a revision request. Declining work, which takes longer than it should because doing it well matters.

Track it for a fortnight and the total is usually between a third and a half of all working time. The best survey evidence on platform work points the same way: the ILO's 2021 World Employment and Social Outlook found workers on online web-based platforms earned on average US$4.90 an hour, falling to US$3.40 once unpaid time was taken into account. The lower end is an experienced earner with templates and a scope line; the upper end is a new one, and a new one also has the slower billable hours.

That figure is the one worth measuring, because it is the one you can move. The billable work is largely irreducible - the thinking is the product - but the overhead responds well to being made boring, which is most of what reusable structure buys you.

Jobs per day, sustainably

There are two numbers here and conflating them is how people burn out.

The first is how many jobs you can do in a day. The second is how many you can do every day for a year.

For a written assessment of any real substance, the burst number is somewhere around eight to ten and the sustainable number is nearer five or six. For recorded work it is markedly lower, because setup, energy and the willingness to be on camera do not scale the way typing does; three to four is a good day and five is a hard one. For live work the constraint is not stamina but the calendar, since a booked slot cannot be moved or resold.

These are ranges from the shape of the work rather than measurements of a population, and yours will differ. What generalises is the ratio: the sustainable number is roughly two thirds of the burst number, for every format, for the same reason.

The reason is that this work is judgement under attention, and attention is the thing that runs out. The fifth assessment of a day is measurably less specific than the first, and specificity is what buyers return for. A day that produces nine perfunctory jobs has traded next quarter's repeat business for this afternoon's income, and the trade is not visible for months. The long-run cost of long weeks is not only commercial: a 2021 WHO and ILO analysis estimated that working 55 or more hours a week is associated with a 35% higher risk of stroke and a 17% higher risk of dying from ischaemic heart disease, compared with working 35-40 hours.

The weekly column already assumes a working pattern of four and a half days rather than five, because the admin has to go somewhere and it is not going to fit in the evenings forever.

Why the week is not five clean days

Two structural facts make a rating week different from a salaried one.

Demand clusters in evenings and weekends, because that is when your buyers are private and unhurried. Being available then is the competitive advantage; it is also precisely the time most people want back, and this is the mechanism by which the trade quietly expands to fill everything.

And the work is bursty at the week level as well as the day level. Three enquiries can arrive within an hour and then nothing for two days. If your capacity plan assumes even arrival, you will alternate between turning work away and having none, which produces the worst of both: a lower average and a higher stress level.

The practical answer is batching. Five jobs in one sitting take meaningfully less time than five jobs across five evenings, because the setup and context-switching costs are paid once rather than five times. That means a stated turnaround measured in days rather than hours, which sounds like a competitive disadvantage and mostly is not: buyers evaluate whether you hit the date, not how short it was. Turnaround as a product feature is the longer version of that argument, and it is the single change that converts a fragmented week into a workable one.

Where the money actually caps out

Put it together and the ceiling has a shape.

Sustainable weekly job count, times your fee, times the weeks you actually work in a year - which is not fifty-two, and is rarely more than forty-four once holidays, illness and quiet periods are counted honestly. Then subtract the platform's cut, the processing charges and whatever your payout route costs, because none of that is yours; the four deductions between a payment and your bank are the difference between the number you quoted and the number you keep.

The result is a real annual figure and it is bounded. This is not a pessimistic reading of the trade; it is the defining structural fact of it, and it is true of every one-person service business from tutoring to translation.

What makes it feel like a surprise is that the first year hides it. In the first year your income rises steeply without any decisions, because your speed is improving and your visibility is growing at the same time - a compounding effect that the piece on why your effective hourly rate keeps moving works through. Both of those flatten, usually within a year, and the flattening reads emotionally as a plateau caused by something you did wrong. It is not. It is the ceiling arriving on schedule.

Scaling means format and price, not volume

Once the hours are full, there are exactly three moves, and one of them is not really available.

Volume is the unavailable one. Beyond the sustainable number, more jobs cost quality, which costs reviews, which costs the price you can charge, which cancels the gain within a couple of months.

Price is available and is the cleanest move, but it needs both preconditions: a record that makes the number plausible and more demand than you can serve at the old one. A rise with only the first is a test of whether anyone stays, and what the first rise actually requires is worth reading before making it rather than after.

Format mix is available, underused, and usually the largest single gain. Earners are consistently wrong about which of their formats pays best per hour, because the one that feels quick is often the one carrying the most unbilled messaging. Thirty jobs of tracked take-home per format answers this definitively, and the answer very often changes what someone offers.

There is a fourth thing that is not scaling but behaves like it: turning work down. Every job taken at the wrong price occupies a slot the right one needed, and at full capacity that is not a philosophical point but an arithmetic one. A backlog is the only real evidence that your rate sits below what the market will bear, and the correct response to a backlog is a price, not longer hours.

The realistic annual shape

A year of this work does not look like twelve equal months.

There is a slow first quarter where the ceiling is evidence rather than hours. There is a middle stretch where speed and visibility compound and income rises without decisions. There is a plateau. And there are two or three genuinely quiet stretches, some seasonal and some not, which are normal and are the subject of how to budget across the lumpy months.

Budget on a trailing six-month average rather than the last month, because the variance is larger than the trend and a strong month is funding for a weak one rather than a raise.

The honest summary is that this is a trade where a part-time week is comfortable, a full-time week is achievable, and a full-time week is genuinely a full-time week, with a ceiling on it that no amount of effort removes. People who do well at it either accept the ceiling and optimise underneath it, or build something alongside it that earns without their presence.

What the ceiling is made of, from outside

Two external facts set the boundaries of the calculation above, and neither is within your control.

The floor of the market is set by automation. Anything a model produces in two seconds is now priced as though a model produced it, and the technical account of what an automated assessment covers is the clearest description of which part of your offer can never be a capacity problem, because it will never be booked.

The demand side sets the shape of the week. Buyers commission a person for judgement and specificity, and the buyer-side account of how someone chooses who to commission explains why they are comparing risk rather than throughput - which is why being visibly busy costs you nothing and being visibly rushed costs you a great deal.

If your product depends on precision, the conventions that make a measurement repeatable are worth being fluent in, because a defensible method is the one part of the work that a faster competitor cannot simply undercut.

For the platform-specific version of capacity - how many jobs a listed judge is expected to take, and how availability is surfaced to buyers - Rate Cock's judges page is the current reference. The general point stands whatever platform you are on: your hours are the inventory, and the only real question is what you charge for a unit of them.