D

Guides

A/B Testing Guide for Growth Teams

A step-by-step guide to running high-quality A/B tests that drive real growth. From hypothesis formation to statistical significance.

1

Define Your Growth Metric

Before running any experiment, identify the one metric that matters most. This should be tied directly to business value — not vanity metrics. Common choices: activation rate, conversion rate, revenue per user, or retention at day 7/30.

Tips

  • Use the ICE framework (Impact, Confidence, Ease) to prioritize which metric to optimize first
  • Ensure you have reliable tracking before running experiments
2

Form a Hypothesis

Write a clear hypothesis: 'If we [change], then [metric] will [improve] because [reason].' The 'because' is critical — it forces you to think about causation, not just correlation. Without it, you're just guessing.

Tips

  • Review heatmaps, session recordings, and user feedback for hypothesis inspiration
  • One hypothesis per experiment — never test multiple hypotheses simultaneously
3

Calculate Sample Size

Before launching, calculate the minimum sample size needed for statistical significance. Use a power analysis with: baseline conversion rate, minimum detectable effect (MDE), significance level (usually 0.05), and power (usually 0.8). Underpowered tests waste time and resources.

Warning

  • Running a test too short is the #1 mistake growth teams make
  • Never peek at results early and stop the test when it 'looks significant'
4

Build and Launch the Experiment

Implement the variation with minimal code changes. Use feature flags to control rollout. Ensure random, even traffic split between control and variant. QA both versions across devices and browsers before going live.

Tips

  • Start with a 50/50 split for maximum statistical power
  • Run tests for full week cycles to account for day-of-week effects
5

Analyze Results and Document

Wait for the pre-calculated sample size before analyzing. Check for statistical significance (p < 0.05). Look at secondary metrics to ensure no negative downstream effects. Document everything: hypothesis, results, screenshots, and learnings.

Tips

  • Build a shared experiment log that the whole team can reference
  • Even 'failed' experiments generate valuable learnings — document them equally

Warning

  • Beware of novelty effects — new designs often show temporary lifts that fade
6

Scale What Works

If the variant wins, roll it out to 100% of users. Then ask: can this learning be applied elsewhere? A winning headline framework might work across multiple pages. A successful pricing change might inform other product tiers. Systematize your wins.

Tips

  • Build a 'growth playbook' of proven patterns your team can reuse
  • Share wins across the company to build experimentation culture

Pro Tips

  • 01Most experiments fail. If your win rate is above 30%, you're not testing bold enough hypotheses — you're optimizing, not learning
  • 02Never trust an A/B test that ran for less than 2 full business weeks. Weekend vs weekday behavior differences will fool you
  • 03The biggest ROI experiments are usually on pricing and packaging, not button colors. Test the things that make your palms sweat
  • 04Build a 'growth playbook' of every experiment — wins and losses. New hires should be able to read it and understand your growth thesis
  • 05If your sample size calculator says you need 50,000 visitors and you get 5,000/month — don't run the test. Redesign for a bigger effect size or pick a different lever
By David Shadrake · Free, no signup required

Strategic Playbooks

Roles That Use This Knowledge

Decisions to Compare

Explore Further

About the Author

David Shadrake

David Shadrake works on strategic business development and tech partnerships, with focus areas across AI, fintech, venture capital, growth, sales, SEO, blockchain, and broader tech innovation. Read more of his perspective on partnerships, market dynamics, and emerging technology at davidshadrake.com.