[ CONVERSION OPTIMISATION ]
A/B Testing Agency
Most A/B tests fail because they are called too early, run on insufficient traffic, or treated as one-off experiments rather than compounding knowledge. We design every test with minimum detectable effect calculations and documented sample-size requirements before a single variant goes live — so you know the test is worth running before you run it.
Statistical Rigour, Not Gut Feels
The industry default is to launch a test, watch the dashboard for a week, and declare a winner when the bar chart looks good. That approach produces false positives at a rate that would concern a first-year statistics student. We calculate the minimum detectable effect, determine the required sample size, and set the runtime before the test launches — not after.
Every test produces a documented learning regardless of outcome. Losing tests tell us something real about your users. That accumulated knowledge compounds — each hypothesis is sharper because of what the last test taught us. We track and share every insight so the learning lives in your business, not just in our spreadsheet.
A/B Testing Services
Test Design & Hypothesis
Every experiment starts with a grounded hypothesis tied to observed user behaviour. We calculate the minimum detectable effect and required sample size before anything goes live — if the test isn't worth running given your traffic, we tell you upfront.
Statistical Analysis & Reporting
Bayesian or frequentist significance testing, depending on your risk tolerance and traffic patterns. No peeking, no early calls. Results include confidence intervals, observed uplift versus predicted, and a plain-English interpretation alongside the numbers.
Copy & Messaging Tests
Headlines, CTAs, value propositions, social proof placement, and microcopy all move conversion rates — often more than layout changes. We write and test copy variants systematically, treating messaging as a first-class experiment variable rather than an afterthought.
Multivariate Testing
When you need to understand how multiple elements interact — not just which variant wins — we design multivariate tests with the right factorial structure and traffic allocation. High-traffic pages where interaction effects matter get the approach they deserve.
Frequently asked questions
How long should an A/B test run?
Long enough to reach your pre-calculated sample size, and never fewer than two full business cycles (typically two weeks minimum). The exact duration depends on your baseline conversion rate, the minimum detectable effect you care about, and your daily traffic. We calculate required runtime before the test launches — stopping early because the variant looks good is the most common cause of false positives in A/B testing.
What can be A/B tested?
Almost any element a user sees or interacts with: headlines, CTAs, hero images, pricing displays, form layouts, checkout flow steps, navigation, social proof placement, page structure, and microcopy. The constraint is traffic — elements deep in a low-traffic funnel take too long to reach significance. We prioritise tests on high-traffic, high-impact pages where results are achievable in a reasonable timeframe.
What statistical significance do you use for A/B tests?
We target 95% confidence (p < 0.05) as a minimum, with 80% statistical power — meaning an 80% chance of detecting a real effect if one exists. For high-traffic, low-risk tests we may accept 90% confidence. For tests with significant business consequences (pricing, major UX restructure) we hold to 95% or higher. We document the chosen threshold and rationale before each test begins, not after the results come in.
How much traffic do we need for A/B testing?
It depends on your baseline conversion rate and the size of improvement you're trying to detect. As a rough guide: if your page converts at 3% and you want to detect a 15% relative improvement (to 3.45%), you need roughly 10,000 visitors per variant. Lower baseline rates or smaller target effects require proportionally more traffic. We run the power calculation for your specific numbers and tell you whether a test is viable before we design it.
Stop Running A/B Tests That Don't Reach Significance
Tell us your current conversion rate and traffic volumes. We'll calculate what's actually testable and which experiments will compound the fastest.
Contacte-nos → Conversion Rate Optimisation →