Testing whether anyone will pay for the work you want to do
Run a cheap demand test on the work you want: priced offers instead of survey interest, what the test costs in hours, and the conversion rate worth acting on.
The advice attached to passion projects is that if you do the work you love visibly enough, the paying version arrives eventually. Sometimes it does. The market it has to arrive from, though, moves for reasons that have nothing to do with what anyone wants to be doing: freelance job posts for writing and coding fell 21% in the eight months after ChatGPT launched, and image-creation posts fell 17% after image generators arrived (Demirci, Hannane and Zhu), while skills referencing AI grew 109% year over year in Upwork's 2026 marketplace data. Demand reprices on its own schedule. Enthusiasm, yours or your audience's, is not a forecast.
Which leaves a narrower and more answerable question. Not "is this a good direction," which nobody can tell you, but "will a specific kind of buyer pay a specific price for a specific version of this, now." That question can be tested in about six weeks for the cost of a slow fortnight's billing, and the test produces a number you can act on.
Interest is free, which is exactly why it predicts nothing
Every validation failure has the same shape: the evidence collected cost the person supplying it nothing. Praise costs nothing. A reply saying the idea is brilliant costs nothing. Signing up to a waiting list costs nothing, which is why waiting lists convert so poorly against their own size.
Evidence is worth what it cost the person who gave it. That is the entire principle, and it sorts every signal you can collect.
Signal | What it costs them | What it predicts | Weight |
|---|---|---|---|
A like, a comment, "I love this" | Nothing | That the phrasing is clear | None |
"Let me know when you launch" | Nothing | That the topic is recognizable | None |
Books a 30-minute call | Half an hour, no exposure | Curiosity | Low |
Describes their own problem, timeline and constraints in writing | 20+ minutes, admits a gap internally | A real problem exists in that role | Moderate |
Asks what it would cost | They have to raise budget internally | Genuine intent — the first real signal | High |
Accepts a written priced offer | Money and internal approval | Demand at that price, from that buyer | Conclusive |
Pays the deposit | Money leaves their account | It survives procurement, which is where deals die | Conclusive, and rarer |
The table is not a ladder everyone climbs. Most conversations stop at row two, and that is fine, provided you count them as row two rather than as encouragement.
What has to be in the test for the result to mean anything
A demand test is not a survey with better manners. It has five properties, and dropping any one of them produces a result you cannot read.
A named buyer, not a market. "Small businesses" is not a buyer. "Operations lead at a 20 to 60 person ecommerce brand that has just taken on a second warehouse" is a buyer, because you can list twelve of them by name and they share a problem. If you cannot list twelve, the test has no sample.
A price on the offer. An offer without a number tests politeness. The price does not have to be right, since being wrong about price is one of the useful results the test can return, but it has to exist and be written down.
A deadline. An offer that stays open forever never gets a decision, and a non-decision is unreadable. A validity date converts silence into a "no," which is data. Both a proposal PDF you can build in the browser and a quote generator with a valid-until date run locally with no signup, no email and no watermark, which matters when you are producing twenty of these for work you may never do again.
Strangers. People who know you buy for reasons that will not scale, and their yes tells you about the relationship rather than the demand. Aim for at least three quarters of the sample to be people with no prior obligation to you. Finding them is its own problem; the good work is never posted, and here's where it moves instead covers where those conversations actually come from.
A small first engagement. The offer should be the smallest complete thing that solves something real: an audit, a pilot, one workshop, one component built. Large offers fail for reasons unrelated to demand, mostly procurement, and then you will misread a budgeting problem as an absence of interest.
What the test costs, so you can compare it to the alternative
Price the test the way you would price a project, because it consumes the same hours. It is the same accounting that prices passive income in billable hours forgone, and it is the only honest way to compare either against client work.
Twenty conversations, at roughly 75 minutes each including research, the call and the written follow-up, is 25 hours. At an illustrative €70 an hour of effective billing, that is €1,750 of revenue not invoiced. Add a few hours for building the offer itself and call it €2,000.
Set that against the alternative, which is not "do nothing." It is committing to a direction on the strength of enthusiasm, spending three to six months repositioning, and finding out from an empty pipeline. At the same €70 an hour, three months at half your normal billing is roughly €18,000 of forgone income. The test is worth running at ten times its price, and that ratio is what makes it worth doing properly rather than as a few casual coffees.
One caution on the denominator. Twenty conversations is a small sample, so treat the result as a signal with wide error bars, not a finding. It is enough to separate "nobody will pay for this" from "somebody will," which is the decision you are actually making. It is not enough to tell you the optimum price.
The other constraint is calendar. Twenty conversations over six weeks is between three and four a week, or about four hours a week including the writing, which fits alongside client delivery and does not fit alongside a full book plus a launch. Let the test run longer than eight weeks and the early conversations go stale, the market shifts underneath you, and you end up comparing answers given under different conditions. A test that drags is a test you will not act on.
Two smaller disciplines make the result more useful. Ask every "no" one question: what would have had to be true for this to be a yes. The pattern across twenty of those answers is often more informative than the two yeses. And do not send an offer you cannot deliver. An accepted offer is a contract, and the fastest way to convert a validated direction into a bad quarter is to win three of them in a week you had already sold.
Three ways the result comes back, and only one of them means the direction is wrong
The failure modes look similar from the inside and require opposite responses, so read them carefully.
Fewer than 8 of your 20 conversations reach a priced offer. This is the most common outcome, and it is almost never a demand verdict. It usually means you talked to people who do not control a budget, or the offer was not legible enough for them to imagine buying it. Both are fixable without changing direction. Change the buyer or change the description, then run ten more conversations.
Eight or more priced offers, and fewer than one in five accepted. Demand exists, because people asked for a price and that costs them something internally. What is wrong is the packaging, the price, or the size of the first step. Test one variable. Halve the scope of the first engagement, or hold scope and move the price. Not both, because then you learn nothing about either.
Ten or more priced offers and no acceptances at all. Now the direction is genuinely in question, at least for that buyer at that price. Before abandoning it, check whether every offer went to the same buyer type, because a single unrepresentative segment can produce a clean zero.
Then there is the good outcome, which is 2 or more acceptances from 8 offers, a 25% close rate. That is not proof of a business. It is proof that the work is purchasable, which is the only thing this test was ever able to establish.
The price floor that stops a successful test becoming an expensive career
New work almost always prices lower at first, because you have no track record in it and the buyer is absorbing that risk. That discount is legitimate and it is also how people end up trapped: three years later they are doing the work they love at two-thirds of what the old work paid, and calling it a lifestyle choice.
Set the floor in advance. The new work has to clear 80% of your current effective hourly rate on the test engagements, and reach 100% within a stated number of projects. Measure it the same way on both lines of work: fee divided by every hour consumed, including scoping, revisions and admin. A timer alone will not tell you this, since hours without the fee beside them are just hours; that gap is the reason people move from a stopwatch to something that carries the money too, which is what a Toggl alternative where tracked hours turn into paid invoices is for. When the two lines are measured the same way, the comparison stops being about which work you prefer and becomes a number.
Where this fits in the wider sequence, and what changes when a second line of work becomes your main one, sits with the money milestones in a complete guide to going freelance. If the direction you are testing is also a reason to leave a salaried job, run the financial tests in three financial tests to run before you hand in notice before, not after.
The rule: 8 offers, 2 yeses, 80% of your rate
From 20 conversations with a named buyer type, at least 15 of them strangers, over six weeks: you need at least 8 to reach a written offer with a price and a date on it, and at least 2 of those 8 accepted at or above 80% of your current effective hourly rate. That is a 40% offer rate and a 25% close, and it is the point at which the direction has earned real commitment — a quarter of your capacity for two quarters, reviewed against the same numbers.
Below 8 offers, the problem is the buyer or the description, so fix one and rerun. Above 8 offers with fewer than 2 acceptances, the problem is price or package, so change one variable and rerun. Two consecutive tests below the line, with the buyer changed between them, is the answer, and it is worth accepting the first time it arrives rather than the fourth.
Worklyn's proposal tracking shows which of those priced offers were opened and which were never read, so a test that returned silence tells you which kind of silence it was.