Free A/B test template

Free A/B Test Template: Plan, Analyze, and Report Every Experiment

A reusable experiment record that keeps planning, execution, results, and decision quality in one place. Download it as Markdown, PDF, or an editable presentation companion.

By Atticus Li · Updated July 28, 2026

What this template is for

An experiment result is easier to trust when the decision rule existed before the result. This template holds the research question, treatment details, measurement plan, execution history, and conclusion together so a later reviewer can see what actually happened.

It does not calculate a result for you. Use the linked test-duration calculator and post-test calculator when you need statistical calculations.

Experiment name

Problem to solve

Research support

Research strengthens the mechanism. It does not prove the treatment caused a result.

Hypothesis

Control

Variant

Exact change

Primary metric

Secondary metrics

Guardrails

Baseline

Minimum detectable effect

Planned sample size

Planned duration

Allocation

Analysis method

Stopping rule

A/B test record

# A/B test record

## Problem and hypothesis
- **Problem:** Not recorded
- **Research support:** Not recorded
- **Hypothesis:** Not recorded
- **Control:** Not recorded
- **Variant:** Not recorded
- **Exact change:** Not recorded

## Measurement plan
- **Primary metric:** Not recorded
- **Secondary metrics:** Not recorded
- **Guardrails:** Not recorded
- **Baseline:** Not recorded
- **Minimum detectable effect:** Not recorded
- **Planned sample:** Not recorded
- **Planned duration:** Not recorded
- **Allocation:** 50/50
- **Analysis method:** Fixed-horizon frequentist
- **Stopping rule:** Not recorded

## Execution and result
- **SRM / allocation check:** Not recorded
- **Execution log:** Not recorded
- **Control counts:** Not recorded
- **Variant counts:** Not recorded
- **Result:** Not recorded
- **Decision:** Not recorded
- **Learning:** Not recorded
- **Evidence tier:** E1

How to use the record

1. Plan: Write the problem, hypothesis, treatment, metric hierarchy, MDE, sample plan, allocation, and stopping rule before the first user is exposed.
2. Run: Add arm-level counts and a plain execution log. Note traffic, campaign, product, targeting, and instrumentation changes.
3. Decide: Report uncertainty and guardrails with the observed outcome, then state the decision and the reusable learning.