You cannot size a market from video. You can sanity-check one.
View counts measure a creator's reach, not a buyer population. But a market of a given size leaves behavioural fingerprints, and those are checkable.
The short answer: get the number from a source that publishes its methodology, then use public content to check whether the market shows the behaviour a market that size would produce — sustained instructional content, competing paid tools, hiring, and elaborate manual workarounds. Fifteen to twenty sources settles it, and a failed check is far more informative than a passed one.
There is a persistent temptation to turn engagement data into market data. It is understandable — the numbers are right there, they are large, and they arrive faster than any industry report. They are also measuring something else entirely, and building a plan on them is one of the more expensive research mistakes available.
Why view counts do not size anything
The mismatch is structural rather than a matter of precision. Video engagement is a function of a creator’s distribution, the platform’s recommendation behaviour, and the entertainment value of the topic. Buyer population is a function of none of those.
The distortion runs in both directions, which is what makes it dangerous. A visually demonstrable hobby topic can generate enormous view counts against a tiny commercial market. A large, well-funded professional category can generate almost nothing, because its practitioners read vendor documentation and attend closed industry events rather than watching tutorials. Neither error corrects with a bigger sample.
| Signal | What it measures | Use for sizing? |
|---|---|---|
| View counts | Creator reach and topic entertainment value | No |
| Subscriber numbers | Channel loyalty, accumulated over years | No |
| Search volume for the workflow | Interest, including students and tourists | Weak, and only as a direction |
| Sustained instructional content | Someone finds it worth teaching repeatedly | As a floor test |
| Competing paid tools with support channels | Money already changing hands | As a floor test |
| Job listings naming the workflow | Organisations paying salaries for it | As a bottom-up input |
The last row is the closest public content gets to a real number, and it is countable rather than inferred: roles that name the workflow imply organisations doing it at a scale worth hiring for. That is a bottom-up input you can defend, unlike anything derived from engagement.
The behavioural fingerprints of a real market
A market where people spend money leaves traces that a market existing only on a slide does not. Checking for them takes an afternoon and catches the most expensive category of error.
- ✗Lots of commentary, no instructional depth
- ✗No paid tools, or one abandoned free one
- ✗Nobody hiring for the workflow by name
- ✗Enthusiasm without any described spend
- ✓Instructional content published continuously for years
- ✓Several paid tools with active support communities
- ✓Job listings naming the workflow explicitly
- ✓People maintaining elaborate manual workarounds
The workaround signal is the strongest of the four. When practitioners build and maintain a spreadsheet, a script or an internal tool to survive a problem, they have already paid for a solution in labour — the clearest demand evidence available short of revenue, and the same signal that makes instructional content productive in finding micro-SaaS ideas in tutorial videos.
Running the check against a number you already have
Start from the sizing you intend to use, wherever it came from, and ask what would have to be true. If the claim is that a hundred thousand organisations do this work, some consequences follow: there should be multiple viable paid tools, a steady stream of practitioner content, and visible hiring. Then go and look.
Three outcomes are possible, and all three are useful. The signals match, and the number survives. The signals are far weaker than the number implies, which means the number is either wrong or is counting people who do not behave like a market. Or the signals are much stronger than the number implies, which usually means the sizing was scoped too narrowly and the real category is larger than the label suggested.
Discovering in week one that a claimed market has none of the fingerprints of spending saves months. It belongs in the same family of early stop signals as the ones in how to disqualify a SaaS idea early — cheap to run, and decisive far more often than founders expect.
Sizing the reachable slice instead
The number that actually governs an early-stage plan is not the total market. It is how many people you can reach in the next six months, and that one is genuinely estimable from public research — because reachable means findable, and findable is what a corpus maps.
If a segment congregates around a dozen channels, three communities and two annual events, you can count those and reason about the population behind them. That figure is smaller and less flattering than a total addressable market, and it is the one that predicts whether your first hundred users exist. It falls out naturally once you have drawn the segment boundary described in defining your ICP from video research.
The one number you can build yourself
Between an unsourced headline figure and no figure at all sits a third option that early-stage founders under-use: a bottom-up count assembled from units you can actually enumerate.
The method is unglamorous. Pick a countable proxy for the buyer — registered businesses in a category, licensed operators, organisations posting a particular role, members of the professional bodies your research surfaced — count what is genuinely countable, then state the multiplier you applied and why. The result is smaller than a top-down figure and infinitely more defensible, because every step is visible and any reader can dispute a specific link rather than the whole chain.
Public research contributes the inputs rather than the arithmetic. It tells you which job titles do this work, which professional bodies and events the segment belongs to, and which tools they pay for — each of which is a countable population somewhere. A sentence like “we identified roughly four thousand organisations posting this role in the last year, of which we estimate a quarter run the workflow in-house” does more for your credibility than any billion-dollar total, precisely because a reader can argue with the quarter.
Where the headline number should come from
For the sizing itself, use sources that publish methodology: government statistical agencies, trade associations that survey their members, regulatory filings, or a bottom-up count you build from countable units and can show your working on. A figure whose provenance you cannot describe in one sentence will not survive its first serious reader, which is the standard applied in writing a market memo you can actually defend.
Pairing a properly sourced number with a behavioural check is a stronger position than either alone. The number gives scale; the check gives confidence that the scale describes people who act like buyers rather than a demographic that happens to exist.
What the pass involves
Fifteen to twenty sources across instructional content, tool comparisons and category commentary, plus an hour with job listings and tool directories. As of August 2026 that fits the $19 a month plan with 25 videos and 2 projects; $59 covers 80 videos and 8 projects for checking several categories, and $199 covers 250 videos, 20 projects and 3 seats. Details are on the pricing page.
Map what a category actually publishes, sells and hires for — with sources attached — before you build a plan on a figure from a slide. 7-day free trial.
Closing thought
Market sizing goes wrong when a number becomes an assumption nobody revisits. The check does not replace it; it just makes sure the world behaves as though the number were true, which is a question worth answering before the number appears on a slide of your own.
Frequently asked
Can you estimate market size from YouTube data?
No, and attempts to do so are where this kind of research goes badly wrong. View counts measure creator reach, not buyer population. What public content can do is sanity-check a sizing you got elsewhere — confirming whether the market shows the behavioural signatures a market that size would have.
What is a sanity check as opposed to an estimate?
An estimate produces a number. A sanity check asks whether a number you already have is consistent with observable behaviour. If a market is supposedly enormous but nobody publishes tooling, tutorials or job listings for it, the number and the world disagree, and the number is usually the problem.
Which public signals actually correlate with a real market?
Sustained instructional content, competing paid tools with active support channels, job listings naming the workflow, dedicated conferences or meetups, and people maintaining elaborate manual workarounds. None of these give you a number; together they tell you whether the market has a floor.
Why are view counts misleading?
Because they measure a creator's distribution, not a category's population. A charismatic channel can put a million views on a niche with three hundred real buyers, and a large professional market can generate almost no video content because practitioners read documentation instead.
Where should market size numbers actually come from?
Industry reports that publish their methodology, government statistics, filings, and bottom-up counts you build yourself from countable units — registered businesses, job postings, or licensed operators. Anything without a stated method is a number someone made up carefully.
What does a failed sanity check look like?
A large claimed market with no instructional content, no competing paid tools, no hiring, and no visible workarounds. That combination usually means the category exists on a slide but not as a group of people who spend money on it.
What does running this cost?
As of August 2026, plans are $19 a month for 25 videos and 2 projects, $59 for 80 videos and 8 projects, and $199 for 250 videos, 20 projects and 3 seats. A sanity check is a small pass — fifteen to twenty sources — and fits the entry tier.