Growth & CRO · Growth & Optimisation

A/B testing with statistical discipline

An experimentation programme with statistical discipline. Hypotheses, MDE, stopping rules and honest reporting even when a test hasn't won.

What it is for

The problem we solve.

Badly run A/B testing doesn't just fail to help: it does damage. Making decisions based on "winners" without statistical significance, without a calculated sample size or without measuring the effect on secondary metrics is worse than not testing at all.

At Hi Labs we treat every test as a scientific experiment: a clear hypothesis, a defined MDE, agreed stopping rules and honest reporting. When a test is inconclusive, we say so. When it is conclusive, we stand by it.

What is included

  • Hypothesis design and prioritisation (ICE / PIE)
  • Implementation with VWO, AB Tasty, Optimizely or GrowthBook
  • Sample size and duration calculation per test
  • Cross-browser, cross-device and cross-funnel QA
  • Reporting with confidence intervals, not misleading percentages
  • Post-rollout impact tracking
How we do it

Our process for A/B Testing.

01

Hypothesis

Every test starts from a clear hypothesis supported by research or prior data.

02

Design

We calculate the sample, define the primary metric and stopping rules before launching.

03

Launch

Thorough QA, gradual rollout and real-time incident monitoring.

04

Decision

Rigorous statistical analysis, a clear recommendation and a roll-out plan.

What sets a sustained CRO programme apart

It isn't the tool. It's the discipline: one hypothesis per test, one primary metric, documentation of every experiment (including the ones that failed) and a repository of learnings that grows over time.

For mid-traffic clients we recommend a cadence of 2-4 tests per month. For high-traffic ecommerce, 8-12. The key is never to test more than you can analyse with rigour.

Contact

Shall we talk about your next project?

Tell us what you want to achieve. We reply within 24 working hours with an honest first assessment.