Freeze the decision rule
before the result moves.
Plan one redacted experiment. For a classic fixed-horizon binary test, the tool estimates an approximate equal-allocation sample and requires both the sample clock and the business-cycle clock. For any sequential or Bayesian engine, enter that platform's own targets instead of reusing the fixed-horizon formula.
Experiment clock and preregistration ledger
| Local version | Type | Created | Hash | Sample / days | Reason |
|---|
The sample estimate is an approximate planning calculation for one equal-allocation, two-arm, two-sided fixed-horizon binary test. It is not valid for arbitrary sequential/Bayesian engines, ratio or revenue metrics, clustered assignment, multiple comparisons, interference, repeated measures or a final statistical decision.
Experimentation and privacy boundary: inputs and the append-only convenience ledger stay in this browser; local storage is user-editable and is not tamper-proof. The paid report receives only the redacted current contract and local version receipts. No analytics, feature-flag, identity or event system is connected. The tool does not test instrumentation, diagnose sample-ratio mismatch, correct multiplicity, analyze outcomes, permit peeking, stop an experiment or declare a winner. An authorized product/data owner must choose the engine and verify its current method.
$9/mo Experiment preregistration membership
Unlock repeat editable Markdown review packs and up to 30 local version/change receipts per experiment, with assumptions, sample and cycle clocks, hashes, exceptions and human signoff. This does not replace your experimentation platform.
- fixed-horizon binary planning assumptions
- sample clock plus business-cycle clock
- external-engine target separation
- append-only local version and change receipts
- owner, analyst and launch-decision signoff
Strong free calculators and why method-specific rules matter
PMF pre-judgment 74/100 · real early-stop governance pain, crowded free calculator market