Solution · Advertising & growth
A/B testing and conversion optimisation
Doubling the conversion rate of a page has the same effect as doubling the advertising budget, at a fraction of the cost. Conversion optimisation looks for the reasons visitors leave without acting and fixes them. A/B testing is how we check that a change actually helps rather than relying on opinions.
[ 01 ]
In short
In short
- 01
Start with data: where do visitors drop out, and what do they do before leaving?
- 02
Each test checks one hypothesis with a clear expected effect on a business metric.
- 03
A test needs enough traffic and enough time; stopping early on a "promising" result is the most common mistake.
- 04
Low-traffic sites benefit more from research and best-practice fixes than from formal A/B tests.
- 05
The winning variant is built properly into the site, not left running inside a testing tool.
[ 02 ]
Scope
Finding what to test
- Analytics: funnel reports show where visitors drop off — product page, cart, form, payment.
- Session recordings and heatmaps: where people click, how far they scroll, where they hesitate.
- Form analytics: which fields cause people to abandon.
- Customer feedback: what sales and support hear — the objections and questions that stop people.
- Speed and mobile usability: slow or awkward pages lose people before content matters.
Forming hypotheses
A good hypothesis states a change, a reason and an expected effect: "Showing delivery cost on the product page will reduce cart abandonment, because customers currently leave at the delivery step when they see the total." We prioritise hypotheses by expected impact, confidence and effort.
Typical things worth testing
- The headline and first screen: does it say what you offer and for whom?
- The offer: price presentation, guarantees, bonuses, urgency.
- Forms: number of fields, order, labels, error messages. See website forms that convert.
- Calls to action: wording, position, contrast, alternatives such as WhatsApp.
- Trust elements: reviews, certificates, project photos, clear contact details.
- Checkout steps and payment options.
Running tests correctly
- Sample size is calculated in advance from current conversion rate and the smallest effect worth detecting.
- Tests run for full weeks, to cover differences between weekdays and weekends.
- One primary metric is chosen before the test — usually leads or purchases, not clicks.
- Traffic is split randomly and consistently, so the same visitor sees the same variant.
- Results are checked for statistical significance before a winner is declared.
The concept and pitfalls are explained in A/B testing.
When traffic is too low
If a page gets a few hundred visits a month, a formal A/B test may take longer than the business can wait. In that case we rely on qualitative research, best practices and before-and-after comparisons over longer periods, and we say clearly that the evidence is weaker.
Tools
Tests can be run with dedicated testing platforms, with feature flags in your own code, or — for ads and landing pages — with the experiment tools of ad platforms. Google Optimize was discontinued in 2023, so tests built on it need a new tool. We choose based on traffic, budget and how the site is built.
What we need from you
- Access to analytics, the website and ad accounts.
- Sales and support input on customer objections.
- Agreement on the primary metric and on how long tests may run.
[ 03 ]
FAQ
Frequently asked questions
How long does an A/B test take?
Until the planned sample size is reached, and at least one or two full weeks. For low-traffic pages, that can be too long to be practical.
Can we test several changes at once?
Multivariate tests are possible but need much more traffic. Most businesses learn faster by testing one meaningful change at a time.
What if the test shows no difference?
That is a result too: the change does not matter much, and effort should go elsewhere. Many tests end this way.