Guide · Programme
How to build an experimentation programme from zero
A few random tests isn't a programme. A programme is a steady rhythm of well run tests that the business trusts. This is the order I set one up in.
1. Make sure you can trust the data
Everything else rests on this. Check that purchases, revenue and the key steps of your funnel are tracked correctly. My list of common GA4 tracking errors is a good place to start.
2. Give it an owner and a goal
Someone needs to be responsible for the programme, even part time. Agree on the one metric that matters most, usually conversion rate or revenue per visitor, so every test points the same way.
3. Fill the backlog with research, not opinions
Use analytics to find where people drop off, then recordings, surveys and reviews to understand why. The ecommerce CRO checklist helps you find the obvious issues fast.
4. Prioritise
Score every idea, for example with ICE, and give confidence points for evidence. Here's how I score test ideas.
5. Agree on a rhythm and a process
Pick a pace you can keep, even if it's one test at a time. Use the same simple steps for every test: hypothesis, design, sample size, build, QA, launch, analysis. A one page test plan template saves a lot of discussion.
6. Decide how you'll decide
Write down the rules before the first test: which statistics you use, the minimum run time, the primary metric, and what counts as a win. It stops arguments later. If you're unsure, start with how significance works.
7. Share what you learn
Keep every result in one place, with screenshots and the decision taken. A short monthly session where you walk the team through wins and losses does more for a testing culture than any slide deck.
8. Show the value
Report conservatively, in money, every quarter. It's what keeps the budget coming. I explain how in how experimentation makes money.
9. Then grow it
Once the process works, bring in more teams, train people to run their own tests, and test bigger things: pricing pages, onboarding, new features. That's when experimentation stops being a project and becomes how the company makes decisions.
Setting this up is what I do in my experimentation consulting: I build the process with your team and coach them until it runs without me.
Want help running tests that give clear answers?
In a free 30 minute call we look at your site and traffic together, and I tell you what I'd test first.
Book a free discovery call