Experiments that ship, with the statistics to back them.
A real experimentation program. Research-grounded hypotheses, pre-registered success metrics, locked sample sizes, and a written memo after every test. Wins or losses - everything is documented.
What a real program ships
What's included
Six deliverables, one program
Research, experiments, tooling, statistical rigor, post-test analysis, and program operations. Take the full program or the pieces that fill your gaps.
Research sprint
Analytics deep-dive, session replay review, funnel breakdown, on-site surveys, and stakeholder interviews. Output: a prioritised experiment backlog.
- GA4 + heatmap review
- 5-7 user interviews
- Prioritised backlog (ICE-scored)
Experiment program
2-4 tests per month, each with a pre-registered hypothesis, success metrics, and a sample-size plan. Wins and losses documented equally.
- Hypothesis + success metric
- Pre-registration doc
- Locked sample sizes
Tool implementation
GrowthBook, Optimizely, LaunchDarkly, VWO, or Convert - we set up the SDK, integrate with your analytics, and train your team.
- SDK + event wiring
- Targeting + audiences
- Team enablement
Statistical rigor
Bayesian or frequentist analysis depending on your traffic profile. Sequential testing where it makes sense. No peeking, no p-hacking.
- Pre-defined MDE
- Sequential / Bayesian analysis
- Power + duration forecasts
Post-test analysis
Every test ends with a written memo - what shipped, what the data said, what we learned, what to test next. Not a Slack screenshot.
- Per-test memo
- Segment deep-dives
- Learnings log
Program operations
Quarterly reviews, experiment calendar, stakeholder digests, and a live backlog your PM and marketing leads can plan against.
- Quarterly program review
- Live experiment calendar
- Stakeholder digest
How it works
Four phases to a compounding program
A two-week diagnostic, then monthly experiment cadence, with a quarterly program review that keeps the backlog sharp.
How it works
From first call to dashboards your clients trust
Research
Two-week diagnostic. Analytics, replays, funnel breakdown, user research. We leave with a prioritised backlog and a first-batch experiment plan.
Instrument
GrowthBook or your preferred tool. SDK wired to your analytics, audiences and targeting configured, team trained on how to ship and read tests.
Ship
2-4 tests per month. Each with a pre-registration doc, locked sample size, and a post-test memo. Wins go to production, losses become learnings.
Compound
Quarterly review of cumulative lift, program health, and backlog refresh. Winners stack, hypothesis-space narrows, unit economics improve.
FAQ
Questions we hear often
What marketing and growth leads ask before committing to an experimentation program.
Ready to ship tests that actually compound?
Book a 30-minute discovery call. We'll walk you through a typical program and show you what the first two months look like.