Growth experiments

Create a structured experimentation system that prioritises high-impact tests, produces valid results, and builds institutional knowledge with every cycle.

Chapters17 Tools5 About

Chapters

Article2 min
A losing experiment is a win if it kills a bad idea cheaply
A clean negative result buys certainty about an idea cheaply. The only failed experiment is one that teaches nothing, win or lose.
Article2 min
An experiment is a bet you wrote down before you placed it
A real experiment is a written prediction tied to a number and decided in advance, so the result can prove you wrong. A tweak without one is just activity.
Article2 min
Build a learnings log so the system never relearns a lesson
Keep every result in a durable log so the team's knowledge of buyers and funnel compounds test by test instead of resetting when people leave.
Article2 min
Common failures
The usual ways experiments go wrong, from calling a tweak a test to changing several things at once, and the habit that prevents each one.
Article2 min
Decide what a win means before the test goes live
Write the success threshold down before launch. Once a number comes back, someone will always argue it is good, so the bar has to exist first.
Article2 min
FAQs
Short answers on running experiments, such as how many to run at once, which is only as many as you can keep isolated and read cleanly.
Article2 min
Most of your tests should be cheap, fast, and small
Growth compounds from a high volume of small, cheap, fast tests. Big swings have a place, but a programme built on them is slow and fragile.
Article5 min
Rank the backlog by expected value, not by enthusiasm
Design experiments with proper controls so you know what caused the change, not just that something changed.
Article2 min
Run it long enough to trust it, and never peek to a stop
Conversion data is noisy early on. Set the run length in advance and never stop a test the moment the number crosses the line, or winners will be false.
Article4 min
Building your backlog
Random testing wastes time and teaches you nothing. Learn how to collect experiment ideas systematically and prioritise them based on potential impact so you always know what to run next.
Article4 min
Creating strong hypotheses
Most experiments fail before they start because the hypothesis is vague or untestable. Learn how to write hypotheses that are specific enough to prove or disprove and tied to metrics that matter.
Article5 min
Setting up experiments
A winning test means nothing if the setup was flawed. Learn how to configure experiments properly in VWO, ad platforms, and email tools so your results are actually valid.
Article5 min
Analysing and acting on results
Statistical significance is just the beginning. Learn how to interpret results correctly, avoid false positives, and turn winning experiments into permanent improvements across your growth engines.
Article3 min
Run your monthly experiment cycle
Run a fixed four-week experiment cycle: pick tests from your backlog, launch them with a checklist, leave them alone, close on a set date and write down what you learned.
Article3 min
Reprioritise backlog by impact and effort
Run a one-hour re-scoring session each quarter that archives stale ideas, updates impact, confidence and effort, and gives you a ranked slate of experiments for the next three months.
Article3 min
Compound learnings across experiments
Turn your experiment log into a short list of beliefs with a confidence level, and check it before every new test, so each result makes the next one better.
Article3 min
Scale winning experiments across channels
Turn one validated experiment win into a set of adapted tests on other channels, ranked by expected value, with realistic targets and a record of what carried over.

Tools

Tool
Heap
Automatically captures every user interaction on web and mobile apps without pre-defining individual events.
Product analytics
Free plan
Sign up on Heap
Tool
Mixpanel
Analyse user events to understand funnels, retention, flows and cohort behaviour without writing SQL.
Product analytics
Free plan
Sign up on Mixpanel
ToolClient acquisition
OmniConvert
A/B testing, web personalisation and on-site surveys all combined in one ecommerce-focused platform.
Growth experiments
Free plan
Sign up on OmniConvert
ToolClient acquisition
Tally
Build forms by typing, with unlimited submissions on the free plan, no payment required.
Marketing funnel
Free plan
Sign up on Tally
ToolClient acquisition
VWO
Runs A/B and multivariate tests on websites to track visitor behaviour and optimise conversion rates.
Growth experiments
Sign up on VWO

About this playbook

I have been testing since 2010, before most B2B companies even knew what a growth experiment was. I am CXL certified in experimentation, I built the testing programme for NU.nl (the largest news site in the Netherlands), and I have run A/B tests for dozens of brands across SaaS, e-commerce, and lead generation. Most companies either skip testing entirely or run random experiments without a clear plan. Both waste time. The ones that grow fastest have a simple system: a backlog of ideas ranked by potential impact, a clean process for running tests, and a habit of documenting what they learn so every experiment builds on the last. This playbook gives you that system. You will learn how to collect and prioritise experiment ideas, write hypotheses that actually teach you something, set up tests properly, and turn results into repeatable wins. No statistics degree required, just discipline and curiosity.