A/B Test Plan template thumbnail

PRODUCT MANAGEMENT TEMPLATE

A/B Test Plan Template

Design an experiment with hypothesis, variants, success metrics, and sample size requirements.

Use this template

What's inside

Field

Value

Name

Descriptive experiment name

Status

Draft

Owner

Person responsible for this experiment

Start Date

End Date

Hypothesis

State your hypothesis using the following structure:

If we [describe the change you are making], then [metric] will [improve/increase/decrease] by [expected magnitude], because [rationale grounded in user behavior, data, or research].

Context / Background

Describe why you are running this experiment. Reference prior data, user research, or business context that motivated the hypothesis.

  • What user behavior or metric prompted this experiment?

  • What have you tried before? What were the results?

  • What is the current baseline performance?

Variants

Variant

Description

% Traffic

Control (A)

Describe the current experience — no changes

e.g., 50%

Variant B

Describe the change being tested

e.g., 50%

Variant C (optional)

Describe an additional change, if testing multiple approaches

e.g., 33%

Primary Metric

Define the single metric that will determine whether this experiment is a success or failure. Be specific about how it is measured and what constitutes a meaningful change.

  • Metric: Name and definition

  • Current baseline: Current value

  • Minimum detectable effect: Smallest change that would be practically meaningful

  • Direction: Increase / Decrease

Secondary Metrics

Metric

Expected Direction

Why It Matters

Secondary metric 1

Increase / Decrease / No change

How it relates to the primary metric or user experience

Secondary metric 2

Increase / Decrease / No change

How it relates to the primary metric or user experience

Secondary metric 3

Increase / Decrease / No change

How it relates to the primary metric or user experience

Guardrail Metrics

List metrics that must NOT degrade as a result of this experiment. If any guardrail metric moves in the wrong direction, investigate before declaring success.

  • Page load time must not increase by more than X%

  • Error rate must remain below X%

  • Customer support ticket volume must not increase

  • Revenue per user must not decrease

Sample Size & Duration

Parameter

Value

Minimum detectable effect (MDE)

e.g., 5% relative improvement

Statistical significance level (alpha)

e.g., 0.05 (5%)

Statistical power (1 - beta)

e.g., 0.80 (80%)

Required sample size per variant

Calculate using your MDE, alpha, and power

Current daily traffic (eligible users)

Number of users who will enter the experiment per day

Estimated experiment duration

Required sample / daily traffic

Targeting / Segmentation

Define who is eligible for this experiment and any segments you plan to analyze separately.

  • Eligible audience: All users / specific plan tier / specific geography / new users only

  • Exclusions: Users to exclude (e.g., internal employees, users in other active experiments)

  • Segments for post-hoc analysis: e.g., new vs. returning users, mobile vs. desktop, plan tier

Implementation Notes

Describe what engineering needs to know to implement this experiment.

  • Where the experiment code should live (frontend / backend / both)

  • Feature flag or experimentation platform configuration

  • Any tracking events that need to be added or modified

  • Edge cases to handle (e.g., users who switch devices, users who clear cookies)

  • How to handle users who have seen the experiment before (sticky assignment)

Results

Complete this section after the experiment reaches the required sample size.

Metric

Control

Variant B

Relative Change

p-value

Significant?

Primary metric

Value

Value

% change

p-value

Yes / No

Secondary metric 1

Value

Value

% change

p-value

Yes / No

Secondary metric 2

Value

Value

% change

p-value

Yes / No

Guardrail metric 1

Value

Value

% change

p-value

Pass / Fail

Summary of findings: Describe what happened and whether the hypothesis was supported.

Decision & Next Steps

Decision

Details

Ship variant?

Yes — ship to 100% / No — revert to control / Iterate — run a follow-up experiment

Rationale

Why this decision was made, including any caveats

Follow-up actions

Next experiment, feature refinement, or broader rollout plan

Decision maker

Who made the call

Decision date

Other Product templates