Prioritising Landing Page Tests by Expected Value

Every optimisation backlog grows faster than any team can clear it. The teams that make progress are not the ones with the most ideas, they are the ones with a consistent way of deciding which idea to build first.

Estimate three things per idea

For each candidate, estimate the share of traffic it affects, the size of the improvement you would expect, and the confidence you have in that expectation. Multiplying the three produces a rough expected value that is good enough for ordering. The precision matters less than applying the same method to every idea.

Be honest about confidence

Confidence should reflect evidence, not enthusiasm. An idea supported by session recordings and a funnel report deserves more confidence than one supported by a colleague’s opinion. Writing the reason for the confidence score next to the score prevents the numbers drifting upward over time.

Cost the build, not just the idea

Divide expected value by the engineering effort required. A moderate improvement delivered in two days often beats a larger one that needs three weeks, especially on a site where the release process is slow.

Clear the easy structural problems first

Before running any experiment, remove the things that are simply broken: a slow checkout step, a form that loses data on error, a mobile layout that hides the call to action. These are not tests, they are fixes, and they should not occupy the backlog that experiments use.

Review the backlog monthly

Re-score the top of the backlog every month. Traffic mix, product changes and prior results all move the estimates, and a list scored once tends to preserve priorities that are no longer correct.