How the novelty effect works
Returning visitors build a habit around a site or an ad. They stop reading the navigation and go straight to the link they always use; they scroll past the banner they have seen for months. Change something visible and that habit breaks. People look, and looking produces clicks.
Those extra clicks are real, but they are a response to unfamiliarity, not to the design being better. Once the new layout becomes the expected one, attention returns to normal and the lift fades. The same mechanism runs in reverse: a redesign that people have to relearn can dip at first and recover as the habit reforms, which is sometimes called a change-aversion effect.
The size of the distortion depends on how much of your audience already knows you. A brand with a large base of returning customers and remarketing traffic carries a lot of habit; a page reached almost entirely by new search visitors carries very little, because nobody has an old version to be surprised by.
Why the novelty effect matters
It is one of the main ways a well-run test produces a wrong decision. The experiment is set up properly, the maths is sound, the winner is declared — and the gain quietly disappears over the following weeks while the team looks elsewhere for the cause.
It matters most when the change is highly visible: a new colour scheme, a repositioned call to action, an unexpected ad format in a feed. It matters least for changes users never consciously notice, such as faster loading or a corrected form validation message. Knowing which sort of change you are testing tells you how suspicious to be of an early lift.
Where the novelty effect goes wrong
The classic error is a short test on a site with loyal repeat visitors, read at the peak of the surprise. The result looks strong, the change ships, and next quarter’s figures are flat. Because the fade is gradual, nobody connects the two.
A second version appears in paid social. A fresh creative format earns attention simply for being unlike the rest of the feed, so it opens well and then declines as the audience becomes used to it. That decline gets blamed on creative fatigue, when the honest reading is that the opening figures were never the steady state.
The third mistake is assuming novelty always inflates. On a familiar interface it can just as easily depress the new version while people relearn where things are, and a good change gets rejected.
What to do about it
Run long enough for the surprise to wear off, then look at the shape of the result rather than its total. Split the period into weeks and compare: a genuine improvement holds roughly level across the run, while a novelty lift is front-loaded and tapers. Extending the planned test duration is usually cheaper than shipping a change that does nothing.
Then separate new visitors from returning ones. New visitors have no previous version to react to, so their behaviour is the cleaner signal; if the lift exists only among returning users, treat it as novelty until proven otherwise. Where a change matters enough, hold a small group back after roll-out and check that the effect is still there once the design is no longer new. That habit is what makes conversion rate optimisation compound instead of resetting each quarter.
