In short
Rarely in the direct sense, because they appear as somebody is already leaving. The costs sit elsewhere: a discount handed to shoppers who were going to buy anyway, an interruption that fires too early on mobile where exit intent cannot really be detected, and a layout shift or slow script that damages the page for everyone. Measure the net effect on revenue per session with a holdout rather than counting signups, and the answer for your store will be clear within a few weeks.
The direct conversion cost is usually small
A properly triggered exit-intent overlay fires when the cursor moves towards the browser controls, which means the visitor had decided to go. Interrupting somebody on their way out risks little of the session that was already ending.
That is the theory, and it holds when the trigger is genuine. It stops holding when the popup is set to appear after a few seconds, on the first scroll, or immediately on landing, which is how a large share of them are actually configured. At that point it is no longer an exit popup, it is an interstitial, and it interrupts people who were reading.
The other direct cost is the ones that fire in the wrong place. An overlay appearing over the cart, during checkout, or on the order confirmation page interrupts a transaction rather than an exit, and those are the placements worth excluding first.
So the honest answer is that the trigger and the placement decide it. The mechanism is not the problem; the configuration usually is.
The expensive cost is discount leakage
An offer shown to everyone leaving is also shown to people who were coming back later to buy at full price. Every one of those redemptions is margin given away for an order you already had.
That leakage is invisible in the popup tool's own reporting, which counts signups and attributed revenue and has no way of knowing what would have happened otherwise. A tool reporting a large attributed revenue figure is often reporting orders that were going to arrive regardless.
Reduce it with suppression rules. Do not show the offer to visitors who have already subscribed, who are logged in, who have bought before, or who arrived from an email you sent. Those groups contain most of the leakage.
Then consider whether the incentive needs to be a discount at all. Early access, a guide, a sizing tool, a restock alert or entry to a draw can capture the address without training every visitor to wait for a code, which is the longer-term cost of a permanent discount popup.
Exit intent barely exists on a phone
There is no cursor to track, so mobile implementations fall back to proxies: a scroll upwards, a period of inactivity, a back gesture, or simply a timer. None of those means the visitor was leaving, which is why mobile popups interrupt far more real sessions than desktop ones do.
That matters because most stores get most of their traffic on phones. A rule set that behaves well on desktop can be doing something quite different to the majority of visitors.
Search engines also take a view on intrusive interstitials that obscure content on mobile shortly after arrival from search. A well-behaved overlay triggered by a considered action is a different thing from a full-screen unit that covers the product as the page loads, and the second is the pattern worth avoiding.
Handle mobile separately. A later trigger, a smaller footprint that does not cover the product, a close control large enough to hit reliably, and a strict frequency cap will keep the mobile version from doing damage that the desktop reporting never shows.
Measure the net effect, not the signups
The only measurement worth acting on is a holdout. Show the popup to a share of visitors and not to the rest, then compare revenue per session between the two groups across the whole period, including the orders that arrive later from the email addresses collected.
That window matters. A popup pays back through the flow that follows it, so a comparison over a few days will undercount its value, and a comparison that only counts immediate orders will undercount it further. Give it long enough to include the welcome sequence.
Also watch page experience. Check that the script is loaded in a way that does not delay the page, and that the overlay does not push content around when it appears, since both affect every visitor rather than only the ones who see it.
Then decide with the whole picture: incremental revenue from captured addresses, minus discount given to people who would have bought anyway, minus any measurable damage to page speed. Stores that run that calculation usually keep the popup and change how it is configured.
A holdout that measures the whole effect
- Group A
- Popup shown, normal rules
- Group B
- No popup at all
- Measured over
- Six weeks, covering the welcome flow
- Primary measure
- Revenue per session, both groups
- Also counted
- Later orders from addresses captured in group A
- Also counted
- Discount redeemed by repeat visitors
- Not used as a measure
- Signup count or the tool's attributed revenue
An illustrative test design. Including later orders on one side and redeemed discount on the other is what turns a flattering report into a decision you can trust.
Related questions
Should the popup offer a discount or something else?
Test both. A discount captures more addresses and trains visitors to expect one, which raises the cost of every future first order. Access, information or a useful tool captures fewer addresses and attracts people more interested in the product, which usually shows up in better flow performance later.
How often should the same visitor see it?
Once, then not again for a long interval, and never after they have subscribed or bought. Frequency capping is stored in the browser, so it will reappear for people who clear their data or use another device. That is acceptable; showing it on every visit is not.
Do popups affect search rankings?
An overlay that obscures the main content on mobile shortly after someone arrives from search can affect how the page is assessed. An exit-triggered overlay on a page the visitor has already read is a different case. Keep the mobile behaviour conservative and the risk stays low.