PM Experiment Examples
(2026 Edition)
Five real-style experiment write-ups — onboarding simplification, streak reminder timing, pricing page redesign, search layout, and paywall placement — each show a specific hypothesis, test setup, primary metric, guardrails, and pre-committed decision criteria. Across all five, hypotheses name an exact metric and target, guardrails catch local wins that cause global losses, and decisions get locked in before results arrive to prevent mid-test rationalisation.
5 real-style A/B test examples with hypothesis, setup, metrics, and pre-committed decision criteria.
By Naman Goyal · Product manager · Builder of PM Streak · Updated July 3, 2026
Build PM Experiment Skills Daily — Free →1. Onboarding simplification
Hypothesis
Reducing onboarding from 5 steps to 3 will lift Day-1 activation from 40% to 50%.
Setup
50/50 split, new signups only, 14-day run
Primary metric
Day-1 activation rate
Guardrails
Day-7 retention (don't lose quality), profile completion (can complete later)
Decision criteria
Ship if activation lifts >5pp AND guardrails within 2pp of baseline
2. Streak reminder timing
Hypothesis
Sending reminders at user's peak activity hour (vs fixed 9am) will lift reminder-driven returns by 30%.
Setup
A/B test on existing users with >3 days of history, 21-day run
Primary metric
Notification-driven return rate
Guardrails
Uninstall rate, notification disable rate
Decision criteria
Ship if return lift >20% AND uninstall/disable rates don't increase
3. Pricing page redesign
Hypothesis
Showing annual pricing first (vs monthly) will lift annual plan mix from 30% to 45%.
Setup
50/50 split, all pricing page visitors, 28-day run
Primary metric
% of conversions on annual plan
Guardrails
Overall conversion rate (don't drop total paid), refund rate (don't upsell unhappy users)
Decision criteria
Ship if annual mix lifts >10pp AND overall conversion doesn't drop >2pp
4. Search result layout change
Hypothesis
Showing 3 products per row (vs 4) on mobile will lift search-to-purchase conversion by 15%.
Setup
A/B test on mobile search results, 14-day run, 50/50 split
Primary metric
Search-to-purchase conversion rate
Guardrails
Session length (don't make browsing annoying), search-to-exit rate
Decision criteria
Ship if conversion lifts >10% without session length dropping >5%
5. Paywall placement
Hypothesis
Moving paywall from Day 3 to Day 7 will reduce free-user churn by 30% without hurting conversion.
Setup
50/50 split on new users, 30-day run to measure downstream effects
Primary metric
Free-user Day-14 retention
Guardrails
Paid conversion rate, LTV (delay doesn't reduce revenue)
Decision criteria
Ship if retention lift >20% AND paid conversion within 5% of baseline
5 Common Patterns Across Great Experiments
Every hypothesis is specific — 'X will move Y from A to B' — not 'this will be better'
Every test has primary + guardrails — protects against local wins that cause global losses
Decision criteria pre-committed — prevents mid-test rationalisation
Run length matches seasonality + statistical power — not arbitrary
Segment considerations noted — new users vs existing, mobile vs web
FAQ
How do PMs write good hypotheses?
A good hypothesis is specific and falsifiable: '[Change] will move [metric] from [baseline] to [target] because [reason].' If you can't name a specific metric or target, the hypothesis is vague. If there's no 'because,' you're not really predicting — you're just hoping.
Should every PM change go through an A/B test?
No. Tests cost time and complexity. Test when: stakes are high, you're uncertain about direction, the change is reversible. Skip tests for: obvious bugs, changes too small to detect, irreversible changes. The judgment of when to test is a PM skill in itself.
Keep learning
Related guides
Practice Experiment Design Daily
Daily PM scenarios on hypotheses, metrics, and decision criteria.
Start Free Trial →