Skip to main content
← Glossary · A/B Testing

Pre-Registration (Experiment Protocol)

The practice of formally documenting an experiment's hypothesis, metrics, sample size, and analysis plan before the test begins, preventing post-hoc rationalization of results.

What Is Pre-Registration?

Pre-registration is the practice of formally documenting your experiment's hypothesis, primary metric, sample size, duration, guardrails, and decision criteria before you launch — and ideally before you've even looked at any preliminary data. It's borrowed from clinical research, where it was introduced to combat p-hacking and publication bias. In A/B testing, it prevents the unconscious (or conscious) cherry-picking of favorable results after the fact.

Also Known As

  • Marketing teams call it the test brief, test plan, or experiment doc.
  • Growth teams say experiment brief or test plan.
  • Product teams use experiment brief, spec, or protocol.
  • Engineering teams refer to it as the test doc or experiment spec.
  • Data science teams call it pre-registration, experiment protocol, or analysis plan.

How It Works

Before launching, you write a one-page document: "Hypothesis: adding a progress bar to checkout will increase completion rate by 5%+ because the Zeigarnik effect creates psychological momentum to finish started tasks. Primary metric: completion rate, measured as checkouts divided by initiated checkouts. Sample size: 40,000 per arm. Duration: 14–28 days. Guardrails: refund rate, AOV, support tickets. Stopping rule: stop when sample size and 14 days reached, or at 42 days. Ship criterion: p<0.05 AND relative lift ≥ 3% AND no guardrail degradation ≥ 2%." Share this with your team, timestamp it, launch.

Best Practices

  • Write the brief in a shared document that captures a timestamp (Notion, Google Docs, or a version-controlled file).
  • Include decision criteria, not just metrics — specify what constitutes ship, don't-ship, and inconclusive.
  • Review the brief at test conclusion against actual results — don't retroactively rewrite.
  • Keep briefs short and structured — one page forces clarity.
  • Build a public experiment log so everyone can see the commitment.

Common Mistakes

  • Writing the brief after seeing preliminary data — defeats the entire purpose.
  • Using vague language ("we'll analyze the results and decide") that provides no pre-commitment.
  • HARKing (Hypothesizing After Results are Known) when the test shows something unexpected and you pretend that's what you were testing.

Industry Context

  • SaaS/B2B: Essential because long tests invite scope creep — pre-registration locks the plan.
  • Ecommerce/DTC: Valuable for keeping merchandising and marketing aligned on test criteria.
  • Lead gen: Often skipped because tests are "obvious" — don't skip; pre-registration catches HARKing even on obvious tests.

The Behavioral Science Connection

Pre-registration leverages the commitment and consistency principle (Cialdini, 1984). Once you've publicly committed to a plan, you're psychologically motivated to follow through. Public commitments are stronger than private ones, which is why sharing the brief with teammates matters.

Key Takeaway

One-page experiment briefs, shared before launch, eliminate more bad decisions than any statistical technique.