A/B testing services

Testing is the part of CRO services that separates what actually worked from what everyone agreed sounded good. It is also the part most often done badly, because a test that is stopped early or sized wrong still produces a confident-looking number.

When is a site ready to test?

When it produces enough conversions, not enough visits.

The number that matters is conversions per variant. Detecting a realistic improvement usually takes hundreds of them on each side, so a page with a few dozen conversions a month cannot answer a question with a test no matter how long it runs. Running one anyway produces noise with a percentage sign on it.

That is not a reason to do nothing. It is a reason to do something else first.

What is worth testing?

Changes large enough to see, aimed at the thing the visitor is stuck on.

Headlines and offers, because they change what the page is about. Form structure, because that is where measurable friction lives. Page layout and where proof sits. Whether prices are shown. The number of steps in a flow. Those produce effects big enough to detect in a sensible time.

Button colours, corner radius and micro-copy on a secondary link do not. The effect, if there is one, is smaller than the noise, so the test reports whatever the random split happened to do that fortnight.

Every test starts from a written hypothesis with a reason behind it, drawn from research rather than from a list of best practices. "We think people leave because the price is not on the page, so showing ranges should raise qualified enquiries" is a hypothesis. "Let us try green" is not.

What counts as a good result?

Better than your own last quarter, first. Benchmarks second.

11.4%

Use that as a horizon, not a target. Categories differ enough that a store and a consultancy can both be doing well at very different rates. The number worth managing is your own trend, measured the same way each month.

Also be realistic about effect size. Most winning tests move things by a few percent, not by double. A programme that reports a large win every month is usually stopping tests early.

Why do so many tests lie?

Four reasons, and all of them are avoidable.

Stopping early. Watching a test daily and calling it the moment it looks significant produces false winners at a rate that surprises people. The sample size is agreed before the switch is flipped, and the test runs to it.

Too many variants. Testing five versions at once multiplies the chance that one of them looks good by accident. Two or three, with a reason for each.

Short windows. A test that runs Tuesday to Thursday measures Tuesday to Thursday. We run whole weeks so weekday and weekend behaviour are both in there.

Measuring the wrong thing. A variant that lifts form starts and lowers qualified enquiries is a loss dressed as a win. We agree the primary metric first, and it is almost always the thing closest to money.

What do you do when there is not enough traffic?

We stop pretending and change the method.

Below the testing threshold, the work is research and rebuilding: recordings, form analytics, the objections your sales team hears, and a clean before and after with an honest statement of how confident we are. That is the same approach used on the landing page optimization side, and on thin traffic it produces better decisions than an underpowered test.

Stores are usually the exception, because order volume gives them the numbers to test properly. The method for those is on the ecommerce CRO page.

Will testing damage our search visibility?

Not if it is set up properly, and it is worth checking rather than assuming.

The failure modes are known. Showing crawlers a different version than visitors is cloaking. A redirect test left running for months starts to look like a permanent move. A variant that loads its content by script can appear empty to a crawler that does not render. Any of those can cost rankings on a page you were trying to improve.

So every test is reviewed against the SEO services plan before it goes live, tests are cleaned up when they finish, and the winner is implemented properly rather than left running in the testing tool forever.

What does the process look like?

What do you get in the report?

The hypothesis, the setup, the numbers and an opinion.

Each test is reported with what we expected, what happened, how confident we are, and what we recommend doing next. Where a result is ambiguous we say it is ambiguous instead of picking the reading that flatters the work.

Over a quarter you get a record of what your audience actually responds to, which is more valuable than any single win. That record is what makes the next year of decisions cheaper, and it is why we keep it in your portal rather than in a tool only we can open. To see how we run this for a team in one of our regional markets, our market page walks through the same process.

How do we start?

Book a free 30-minute consultation and we will show you where the obvious conversion problems are. On most sites there are changes worth shipping before a testing programme is worth starting.

If you already have the traffic, book a call and we will scope the programme. Plans are flat monthly retainers with starting prices on the pricing page, month to month, and you can stop at the end of any month.