Push Notification Opt-In at Deposit Doubles Abandonment 26%

Operators who trigger a push-notification permission prompt at the moment a player opens the deposit screen see 26% more of those sessions end without a completed deposit than operators who ask at registration. That figure comes from a 2023–2024 comparison of 14 mid-size operators across Croatia, Serbia and Romania, covering roughly 2.1 million deposit-screen sessions, and it runs against the assumption most CRM teams still work from — that a prompt shown at the point of highest intent converts best. It does not. It converts the player's attention, briefly, and then costs the operator the deposit.

The mechanism is worth spelling out because the fix is not "stop asking." Push permission on Android and iOS is a one-shot, system-level dialog. The player either grants it or the app never asks again without a trip to settings. Asking at deposit means you are interrupting a task the player has already decided to complete with a decision they have not thought about, on a screen where any friction reads as risk. Croatian players depositing in euros through a bank card or a local wallet are, at that exact second, looking at an amount, a payment method and a balance. A modal from the operating system is not part of that picture.

What the 26% actually measures

The headline number is an abandonment delta, not a conversion rate. The baseline in the study was a control group prompted at first login: 31.4% of deposit-screen sessions in that cohort ended without a deposit. In the test cohort, prompted on deposit-screen entry, the figure was 39.6%. The 26% is the relative increase — 8.2 percentage points on a base of 31.4. That distinction matters when you take the number to a finance meeting, because 26% relative sounds like a crisis and 8.2 points sounds like a Tuesday.

The study's authors controlled for the obvious confounders. Sessions were matched by player tenure band, average deposit size and device OS. Android and iOS behaved differently: the Android delta was 29%, iOS 21%. The gap is consistent with the fact that Android's permission dialog is more visually intrusive and, on some OEM skins, appears with a full-screen scrim. Players on iOS were also more likely to have already denied notifications at the OS level during onboarding, which means a slice of the iOS cohort never saw a prompt at all and could not be affected by it.

One more measurement caveat. The study counted a session as abandoned if no deposit settled within 30 minutes of the deposit screen opening. That window will misclassify a small number of players who completed a deposit later the same evening. The authors estimate the misclassification at under 2% of sessions and say it applies roughly equally to both cohorts, so the delta survives, but the absolute abandonment rates in the study should not be lifted into your own dashboards without re-basing.

Why the effect is larger than it looks

Abandonment at deposit is not evenly distributed. In the study's control cohort, 61% of abandoned sessions ended within 90 seconds of the screen opening. Those are the players the prompt is most likely to catch, because the permission dialog fires on entry, before the player has typed an amount. A player who abandons at 90 seconds has not been defeated by payment friction or a failed 3-D Secure step. They were interrupted at the top of the funnel.

The revenue arithmetic is where this gets uncomfortable. If a mid-size Croatian-facing operator processes 40,000 deposit-screen sessions a month at an average deposit of €47, an 8.2-point abandonment increase removes roughly 3,280 sessions. Not all of those would have converted at the same rate, and some players return later. But even at a 40% recovery assumption, you are looking at something in the region of €60,000 a month in deposits that never happened, on a base where the operator's margin on those deposits is a fraction of the handle. The push list you gained is worth a small fraction of that. This is the trade the study is really describing: a subscription asset bought with deposit revenue, at a bad exchange rate.

Where the prompt should go instead

The study's own follow-up tested three alternative trigger points, and the ranking is stable enough to act on.

After first deposit settles. Prompting on the confirmation screen — after the payment has cleared and the player is looking at a balance that just went up — produced a 41% opt-in rate with no measurable abandonment effect. The player is between tasks. The dialog does not sit between them and anything they want.

On the second session, not the first. Players who returned within 72 hours of registering opted in at 38% when prompted at session start, versus 22% when prompted during registration. Return behaviour is the cheapest available proxy for genuine interest, and it costs nothing to wait for it.

Never on the deposit screen, at any point in the flow. The study tested a softer variant — a pre-permission explainer sheet shown before the OS dialog on deposit entry — and it still produced a 14% relative increase in abandonment. The explainer reduced the damage but did not remove it. The lesson is that the OS dialog is not the only cost; any interruption on that screen carries one.

The pre-permission sheet, used correctly

The explainer sheet is still the right tool, just not there. A two-line in-app sheet that states what the notifications will contain ("deposit confirmations and bonus expiry reminders, no more than four a month") before the OS dialog lifts opt-in rates by 11–19% in most published tests, because it filters out players who would have denied and spares them the one-shot OS prompt. The specific number matters more than the design. "Four a month" is a promise a player can evaluate. "Important updates" is not, and players in this market have been trained by three years of bonus spam to read it as a threat.

Timing within the session also matters less than most teams assume. The study found no significant difference between prompting at session start versus mid-session, provided the player was not mid-task. The variable that moved opt-in was whether the player had something unfinished on screen. Deposit is the most unfinished screen in the product.

The Croatian context changes the calculus

Push opt-in rates in Croatia run lower than the Western European average, and the reasons are structural rather than attitudinal. Croatian players skew toward web over app — the study's Croatian cohort was 58% web, against 34% in the Romanian cohort — and web push permission on Chrome and Firefox is a different animal from the native app dialog. Browser prompts are visually softer, can be dismissed without a decision, and are frequently blocked by the browser itself after repeated denials. The 26% figure is an app-dialog figure. Applying it directly to a web-first Croatian funnel will overstate the effect.

There is also a regulatory dimension that operators in this market tend to handle badly. Under the Croatian Act on Electronic Communications, unsolicited electronic communication for marketing purposes requires prior consent, and the consent has to be specific and informed. A push notification that promotes a bonus is marketing. A push notification that confirms a deposit is transactional. The distinction is not academic: the study's test cohort operators were sending a mixed stream, and the opt-in rate fell by 9 points when players were told the notifications would include offers. Operators who separate the two streams — transactional on by default where the OS allows it, promotional behind an explicit opt-in — get better numbers on both.

What the payment mix does to the funnel

Croatian deposit funnels are longer than the study's average because of the payment mix. Card deposits with 3-D Secure add a step; local bank transfers add a redirect; and the players most likely to abandon at the permission prompt are the same players most likely to abandon at the 3-D Secure challenge. The two effects compound. If you are running a prompt on deposit entry in a market where 40% of deposits route through a redirect flow, the interruption lands before a process the player already finds annoying, and the abandonment you attribute to payment friction may partly be yours.

The practical check is cheap. Pull your deposit-screen abandonment by payment method for the last 90 days and see whether the rate is flat across methods. If card abandonment is materially higher than wallet abandonment, you have a payment problem. If it is flat, and you are prompting on that screen, you have a prompt problem.

What to do with the list once you have it

A push list is only worth the interruption it cost if the messages earn their place. The study tracked opt-out rates over 90 days and found that operators sending more than six promotional pushes a month lost 18% of their list, against 6% for operators sending two or fewer. Transactional messages — deposit confirmations, withdrawal status, session reminders — had opt-out rates under 2% regardless of frequency, which suggests the tolerance is for marketing specifically, not for notifications in general.

The segmentation that mattered most was deposit recency. Players who had deposited in the last seven days responded to bonus pushes at roughly three times the rate of players at 30 days dormant, and the dormant segment's opt-out rate was double. Sending the same push to both is how lists die. If your CRM tool cannot segment on deposit recency at send time, the push list is not an asset yet.

One number worth holding onto: in the study, players who opted in after first deposit settled deposited 1.4 times more often in the following 30 days than matched players who were never prompted. That is a selection effect more than a causation — interested players opt in — but it means the list you build at the right moment is a better list than the one you build at the wrong moment, and it is the same size or larger.

The question the study does not answer

The open question is whether the 26% holds when the prompt is the only thing on the screen. Every test in the study ran the permission dialog against a live deposit screen with an amount field, a payment selector and a balance. Nobody has published a clean test of a deposit screen with the prompt removed and nothing else changed, because operators rarely run that experiment — it feels like leaving money on the table, and the control group is expensive to maintain. But until someone does, the 26% is a comparison between two prompt placements, not a measurement of the prompt's total cost. It is entirely possible that the true figure is higher, and that the operators still prompting at deposit are paying more than they know.

The cheaper experiment is available to any operator reading this. Split your next 50,000 deposit-screen sessions, prompt half at confirmation and half on entry, and measure settled deposits at 30 minutes and at 24 hours. If your delta comes in under 8 points, you are either better at this than the study's cohort or your payment mix is masking the effect. Either way, you will know something the published research does not, and you will have paid for it in deposits rather than in a subscription to someone else's benchmark.