# When to stop optimising a page and move to the next one

> Diminishing returns on a page show up sooner than most teams admit. Four signs that a page has given you what it's going to give, and where the next win is likely to be.

- Source: https://bixlow.com/en/blog/when-to-stop-optimising-a-page
- Language: en
- Published: 2026-11-16T09:00:00.000Z
- Updated: 2026-10-06T09:28:18.400Z
- Author: fernando
- Topics: experimentation program, experiment design, checkout optimization, audience targeting, CRO, minimum detectable effect, new brands
- Publisher: Bixlow

---
Teams get attached to a page. It's the one they know, the one with the traffic, the one where they had their first win, so they keep testing it long after it has stopped being the best place to spend a month. There are four fairly reliable ways to tell when that point has passed.

## The effects keep getting smaller

The first test on the page moved conversion by 18%. The second moved it by 7%. The third was inconclusive. That is what diminishing returns normally look like, and it's the clearest signal you'll get. The big structural problems have been fixed, and what remains are refinements smaller than your minimum detectable effect, so carrying on costs you months for changes you can't measure.

Check it honestly. Write down the last three experiments on that page with their effect sizes, and if the trend is downward and the latest one was inconclusive, the page is finished for now.

## Your remaining ideas are all wording

Look at the list of ideas you still have for the page. If it's all copy variants, moving elements around and colour changes, you've used up the structural hypotheses, and structural changes are the only kind a small brand can detect.

One way to check: can you describe the next hypothesis in terms of something the visitor will be able to do or understand that they couldn't before? If the honest description is "the same thing, phrased better", you're past the useful point.

## Another page has worse numbers and similar traffic

This is simple arithmetic and gets ignored all the time. Say your product page converts at 3.1% after three rounds of optimisation, and your cart loses 68% of the people who reach it. The money is in the cart, even though the product page is the one you know how to improve.

Make a habit of looking at the drop-off across the whole funnel before you pick the next experiment, instead of starting from the page you already had in mind.

## The page is going to change anyway

If a redesign, a platform migration or a change to the product range is coming in two months, don't spend your sample on the current version. What you learn about structure might carry over to the new one, but the specific variants won't.

Use the time to instrument the new version properly before it launches, and record a baseline so you can tell whether the redesign helped. Redesigns get shipped on faith remarkably often, and a before-and-after comparison with no control group is exactly the kind of weak evidence that lets a harmful redesign survive.

## Where to look next

Once a page is exhausted, the next opportunity is usually in one of three places.

Further down the funnel is the most common. Cart and checkout are less optimised than product pages almost everywhere, because people are more nervous about touching them, and for that same reason the effects of checkout optimisation tend to be larger.

It may be a segment rather than a page. You might be done with the page for visitors in general and not done with it for mobile visitors arriving from paid social. Experiments aimed at a specific audience often find effects on a page whose overall numbers have nothing left to give.

Or the problem may sit before the site altogether. If the page is fine and people still don't convert, the mismatch could be in what the ad promised or in who it was shown to. That isn't something page testing can fix, however much of it you do.

## Coming back to it

A page that's finished for now isn't finished for good. One you exhausted at 400 sessions a day becomes testable again at 1,200, because effects that were too small to detect are now within reach.

So keep the hypotheses you shelved for being too small, in a list with a note of why each was parked. When traffic grows meaningfully, that list will be the best-prepared backlog you have: ideas that have already been thought through and can now be measured.

---

## Meet our AI CRO agent

What you just read, put to work by an agent: it understands your visitors, decides what to test or personalize, builds it on your site without a developer and tells you when a result is real.

https://bixlow.com/en/ai-cro-agent
