Free tools for questions you keep guessing at
Can your eval see the regression you are worried about?
Three tools, free, no signup, running entirely in your browser. Answer that question in ten seconds instead of arguing about a three-point delta for an afternoon.
The problem
The questions are simple, the formulas are known, and everybody guesses anyway
How many eval cases do I need. Is this three-point move real. How much does checking every morning cost me. Each has a closed-form answer that takes seconds to compute, and almost nobody computes it — because the tools that exist were built for clinical trials and ask for parameters an engineer does not have.
The insight
Give away the tool, keep the relationship
Every engineer who reaches for a power calculator has, by definition, the problem the paid products solve: they are making a decision on a number they are not sure about. A free tool that answers their immediate question honestly is worth more than a landing page that describes a product, because it is used rather than read — and the person who used it remembers who built it.
Paired and unpaired two-proportion power, minimum detectable effect and required sample size, plus always-valid sequential bounds — the same functions the paid products run, exposed directly.
How it works
Four steps, no data science team
Eval suite, A/B test, or a running experiment you keep checking.
Pass rate and case count. Nothing you need to look up.
Usable or underpowered, with the number of cases that would fix it.
No account, no export gate, no follow-up email.
Who it is for
Free, permanently, with no capture
Any engineer about to make a decision on a number. Deliberately the broadest audience in this portfolio.
Pricing
- –Power and MDE
- –Required sample size
- –Sequential bounds
- –Runs client-side
Competition
What exists, and what it does not do
| Who | What they do | The gap |
|---|---|---|
| Evan Miller and the classic calculators | The tools engineers currently find and use. | Excellent, built for conversion A/B tests, and they do not handle the paired design that an eval suite actually is. |
| G*Power and academic tools | Comprehensive power analysis. | Built for researchers, ask for parameters an engineer does not have, and nobody installs a desktop application for this. |
| Asking a model | Increasingly the default. | Gets the formula right and cannot tell you whether your particular suite is underpowered without you knowing which question to ask. |
| Guessing | The incumbent. | Free, instant, and wrong in a direction that always favours shipping. |
There is no business here and pretending otherwise would be the mistake. The risk is opportunity cost — time spent polishing free tools is time not spent talking to customers — and the mitigation is that this is already built, so shipping it costs an afternoon rather than a sprint. The second risk is that it works: a tool with real traffic creates support expectations for something that will never generate revenue.
Market
None. This is the top of the funnel and should never be priced
Every engineer running an eval. The only metric that matters is whether people arrive at the paid products having already used this one.