Methodology
Last updated July 2026
Every cost figure and performance claim on this site came from somewhere. This page says where, what the sourcing labels mean, and which claims are weaker than they look. It exists so you can check the work rather than take it on trust.
The rule
A number gets published only if it can be traced back to a source, and the source is named on the page carrying it. Where a figure is an estimate, it is labelled as an estimate and the assumptions behind it are printed alongside. Where something has not been tested, the guide says so rather than filling the gap with a plausible figure.
Part of this is structural rather than a matter of editing discipline. Every guide carries fields recording what evidence is behind it, which external sources it draws on and when those were last opened, and those fields are visible on the page rather than kept internally. That does not make a claim true. It does mean an unsourced one has nowhere to hide.
Evidence labels
Guides are tagged with the kinds of evidence behind them. These are the labels in use, in rough order of how much weight they deserve:
- First-hand testing
- We ran it and are reporting what happened. The guide describes the run in specifics (what was produced, over what period, and what it cost) so you can judge whether it resembles your situation.
- Provider pricing page
- Quoted from the provider's own published pricing page, with the date it was checked. Strong evidence, and the fastest to go stale.
- Platform policy
- Quoted from published YouTube policy rather than from what people say the policy is. Enforcement in practice can differ from the written rule, and where that is known the guide says so.
- API documentation
- Quoted from official documentation. Reliable for limits and behaviour, and worth less than testing for anything to do with quality or speed under load.
- Public data
- Publicly observable channel or video data, gathered from what the platform shows anyone. Real, but partial: it cannot see revenue, cost, or anything behind the analytics wall.
- Community report
- Reports from forums and threads. The weakest kind here, and always labelled as such: useful for noticing that something has changed, never sufficient on its own for a number.
The last one is worth singling out. Community reports (a thread where several people say a tool started throttling them, for example) are genuinely useful signal and genuinely weak evidence. They are cited where they are the only thing available, and always labelled, so you can weigh them accordingly.
Where cost figures come from
Provider prices are quoted from the provider's own published pricing page on the date given in the guide's last-verified field. They are not averaged, not marked up, and not adjusted for any deal or credit that is not available to everyone.
Any figure describing what a real run cost is measured from the provider's own usage reporting rather than estimated from token counts, because the two disagree often enough to matter. Guides that quote a per-video or per-script cost state what was included and what was left out. An "included costs" and "excluded costs" list is a required part of a cost breakdown, for the obvious reason that a cost figure without a boundary is not a figure.
What is not tested
Model output quality is not benchmarked here. There is no scoring rubric behind statements that one model writes better scripts than another, and where such a comparison appears it is a stated opinion from use, labelled as one. Anyone presenting a quality ranking of writing models as a measurement is overstating what they have.
Nothing on this site measures whether a method grows a channel. Views depend on topic, packaging, timing and audience, none of which a production stack controls. Guides describe what a stack costs and what it produces, and stop short of promising what it will earn.
How often pages are re-checked
API prices move, tools are discontinued, and platform policy changes. Every guide carries a last-verified date and a refresh schedule chosen from this list:
- Monthly
- Anything quoting live API prices.
- Quarterly
- Tooling and workflow guides.
- Every six months
- Platform policy and strategy.
- Annually
- Conceptual pieces that age slowly.
- No schedule
- Evergreen. Revisit only when something changes.
The date shown on a guide is the date its claims were last checked, not the date the file was edited. A typo fix does not reset it. A guide past its refresh date is marked as needing one, and that marking is a deliberate editorial act rather than something a script decides, so if you find a stale figure on a page that has not been flagged, that is a miss worth reporting.
Corrections
When something here turns out to be wrong, the page is corrected and the change is recorded on it, rather than edited silently. Substantive corrections appear in the guide's change log with a date. If you find an error, saying so is genuinely useful.
Send corrections to ethan@buildtuber.com. Pointing at the specific figure and where it went wrong is more useful than a general note, and a correction that lands is worth more here than a compliment.
Conflicts of interest
This site sells products that do some of what the guides teach. That is a conflict worth stating plainly, and it is stated in full on the disclosures page, along with how the site makes money and what it does not do.
Get the build guides as they go out.
One email when a new guide ships. The full method, not a teaser. Unsubscribe whenever.