Whether it offers a newsletter or a discount, a pop-up exists to raise conversions. Sometimes it genuinely does. The trouble is that the increase is visible while what the same pop-up costs elsewhere is not. This piece is about the invisible half.
The cost lands in two places at once
The first is search. When someone taps a result on a phone and finds an interstitial covering the content they were promised, the page can be assessed as one that does not deliver what the result advertised. Search engines have flagged this arrangement for years, and they single out interstitials that obscure content on mobile in particular.
The second is conversion. The sign-up rate from the pop-up is easy to see; the person who saw it and hit the back button appears in no report at all. They did not subscribe, they did not enquire, and statistically they are simply a bounce. Gains are counted and losses are not, which means a pop-up’s scorecard always reads better than its actual effect.
Not every overlay is the same thing
There is an important distinction here. What causes trouble in search is a large panel that covers the content on arrival, not anything that ever appears over a page. Notices that exist for legal reasons — cookie consent, age verification, login prompts — are treated separately when they take up a reasonable amount of space.
That distinction is practically useful because it means you can avoid the problem without removing the pop-up. Change its shape instead: a bottom banner rather than a full screen, after reading rather than on arrival, a comfortable dismiss target rather than a five-pixel cross.
Three levers, and only three
Timing. The most common setting is three seconds after arrival, and it is close to the worst possible one. At that moment the visitor still does not know what the page is, so they have nothing to judge the offer against — and people decline offers they cannot evaluate.
Trigger. Prefer behaviour over elapsed time. Someone who has scrolled past half the article has signalled interest, and an offer at that point reads as a next step rather than an interruption. Exit intent works on desktop, where a cursor leaving towards the top of the window means something, but no such signal exists on a touchscreen — so never use one configuration for both.
Dismissal. The close control has to be hittable with a thumb and must never sit off the edge of the viewport. And once someone has closed it, do not show it again during the same visit. Closing it was an answer; repeating the question is not conversation, it is nuisance.
If you are keeping it, check it properly
The fastest check is to reproduce a real arrival. Search on your phone, tap your own result, and see what happens — it is frequently nothing like the view you get on a desktop while logged in as an administrator. If you are running ads, repeat the test through an ad link. A pop-up on a landing page means you are paying per click to show people an overlay instead of the page.
And when you review the pop-up’s performance, put the sign-up count next to the page’s bounce rate and its next-step completion. Subscriptions up and enquiries down is not a win.
Measuring the relationship between speed and conversion is covered in the speed and conversion series, and the ways a page gets heavy live in the performance archive. To see exactly which scripts — pop-up tooling included — actually load on your pages, that is part of the diagnostic in our optimization program.