Hold Screen at 3s Adds 19% to Withdrawal Confirmations

A 3-second hold on the withdrawal confirmation button lifted completed withdrawal requests by 19% in a controlled test run across four Croatian-facing operators between September and November 2024. The mechanism was not a bonus, a limit change, or a new verification step. It was a single UI intervention: the confirm button required a continuous press for 3,000 milliseconds before the request was submitted. The result matters because withdrawal friction is usually discussed as a compliance cost or a player-experience tax, and almost never as something that can be tuned to a measured number.

The test itself is worth describing precisely, because the 19% figure is easy to misread. Four operators, all holding Croatian licences and serving domestic traffic, each ran an A/B split on their cashier flow for eleven weeks. Control users saw the existing one-tap confirm. Test users saw a press-and-hold button with a 3-second fill animation and a haptic tick at completion. Assignment was per-account, not per-session, so a player stayed in the same arm throughout. The primary metric was the share of withdrawal requests that reached "paid" status within 72 hours of submission. Secondary metrics included request abandonment, support tickets tagged "withdrawal", and re-deposit behaviour within 24 hours of a request.

The headline number is the lift in that primary metric. The secondary numbers are where the interesting argument lives.

What the 19% actually measured

The 19% is a relative lift, not an absolute one. Baseline completion inside 72 hours across the four operators sat at 61.4% in control. Test arms landed at 73.1%. That is a relative improvement of roughly 19%, and an absolute improvement of 11.7 percentage points. Anyone quoting the figure in a board deck should quote both, because 19% relative sounds like a marketing stat and 11.7 points sounds like an operational one. It is the second.

Worth noting what did not move. Median time from request to payment did not change materially — 14.2 hours in control, 13.9 in test. The hold did not speed up the back office. It changed which requests survived long enough to be paid.

That distinction is the whole point. Withdrawal abandonment is not usually a decision players make after deliberation. It is a decision made in the two-to-forty-second window between tapping confirm and the request being locked. In that window a player can still cancel, still reverse, still decide the money would be more useful in the balance. The 3-second hold does not remove that option. It just moves the decision point earlier, to a moment when the player is actively pressing rather than passively glancing at a confirmation screen.

The abandonment split

Request abandonment — the share of initiated withdrawals that were cancelled by the player before reaching the payment queue — fell from 22.8% to 16.1% in the test arms. That is the single largest contributor to the completion lift. The remaining contribution came from a modest reduction in failed requests due to mistyped IBANs and mismatched account names, which fell from 4.1% to 2.9%. Whether the hold deserves credit for that second effect is debatable; the test arms also shipped a slightly clearer account-name field, and the operators could not cleanly separate the two. Treat the IBAN improvement as confounded.

The abandonment number is the one to interrogate. A 6.7-point drop in player-initiated cancellations is large, and there are three plausible explanations, only one of which is flattering.

The first is commitment. Pressing and holding for three seconds is a small act of intent. It makes the request feel like a decision rather than a tap, and players who complete the hold are marginally less likely to reverse it in the following minutes.

The second is friction-as-delay. The hold adds three seconds before the request enters the queue. If a meaningful share of cancellations happen in the first few seconds after submission — a reflex, a second thought, a partner looking over a shoulder — then simply delaying the moment of submission by three seconds could capture some of those reversals inside the hold itself, where they register as an incomplete press rather than a cancelled request. That would flatter the metric without changing behaviour.

The third is instrumentation. If the test arm's cancel button was rendered differently, or moved, or required its own confirmation, then part of the abandonment drop is a measurement artefact. The operators involved say the cancel flow was untouched. Independent verification is not available, and readers should weight the 19% accordingly.

Why the hold is not the same as a cool-off period

Croatian operators already have a familiar tool in this space: the mandatory cool-off between request and payment, which most domestic licensees set somewhere between a few hours and a full day depending on the payment method. That tool is a delay. The hold is not.

A cool-off period inserts time after the request is locked, during which the player can still reverse but must actively navigate back into the cashier to do so. It reduces impulsive reversal by adding steps. The hold inserts time before the request is locked, during which the player is physically pressing a button. It reduces impulsive reversal by adding commitment. The two are complementary, and the test data suggests they stack: the largest completion lifts appeared in the payment-method segments that already had the longest cool-off windows.

There is a compliance argument here that deserves more attention than it usually gets. Under the Croatian gambling framework, operators carry obligations around identifying and responding to problematic play patterns, and withdrawal friction is one of the levers available. A hold is a cheap lever. It costs nothing to implement, requires no regulatory notification in most interpretations, and does not restrict the player's access to their own money beyond a three-second press. Compared with deposit limits, which players often resist, or session timers, which they route around, a hold on the confirm button is a friction the player barely notices and cannot easily disable.

That is also the argument against it, and it should be stated plainly. A friction the player cannot see is a friction the player cannot consent to. Three seconds is trivial. But the principle — that the operator is inserting deliberate delay into a money-out flow to change player behaviour — is the same principle that underlies dark patterns, and the difference between the two is one of degree and disclosure, not kind. Any operator adopting this should label it. A short line under the button — "hold to confirm, this takes 3 seconds" — costs nothing and removes the objection.

Where the effect was strongest

The lift was not uniform. Segmenting the four operators' pooled data:

• Players aged 18–24 saw a 26% relative lift, the largest of any age band. This group also had the highest baseline abandonment, so the absolute effect was largest here too.
• Players with a withdrawal request in the previous 30 days saw an 8% lift. Familiarity with the flow blunts the effect considerably.
• Players requesting under €200 saw a 23% lift; those requesting over €1,000 saw 11%. Small withdrawals are the impulsive ones.
• The effect was strongest on Friday and Saturday evenings, where relative lift reached 31% against a weekday baseline of 14%.

That last number is the one that reframes the whole finding. The hold is not primarily a UX improvement. It is a speed bump positioned exactly where impulsive reversal is most likely, and it works best when the player is least deliberate. Read that way, the 19% average understates the effect on the population that matters and overstates it on the population that does not.

The counter-argument: what 19% costs

Every friction intervention trades one metric for another, and the test data shows the trade clearly. Support tickets tagged "withdrawal" rose 12% in the test arms, concentrated almost entirely in the first two weeks. The common complaint was not the three seconds. It was uncertainty — players pressing, seeing the fill animation stall because their finger slipped, and assuming the request had failed. The fix is obvious and was applied mid-test by two of the four operators: a clear "keep holding" state and an unambiguous success confirmation. Post-fix, the ticket uplift fell to 4%.

The more serious cost is the one the operators did not measure well: whether some players, faced with a hold, simply did not initiate the withdrawal at all. The test captured abandonment of initiated requests. It did not capture the share of players who opened the cashier, saw the new button, and closed it. If that number is non-trivial, the 19% completion lift is partly a selection effect — the hold is filtering out the least committed requests before they are counted, which improves the completion rate without improving anything the player cares about.

The operators' defence is that the cashier open-to-request conversion rate fell by only 1.4 points, from 78.9% to 77.5%, which they argue bounds the selection effect. That is a reasonable argument and an incomplete one. A 1.4-point drop in initiation, applied to a large base, is a real number of players who did not get their money out as quickly as they would have otherwise. Whether that is a cost or a benefit depends entirely on why they stopped, and the test cannot tell us.

There is also a simpler objection. The 19% figure comes from four operators over eleven weeks. That is a small sample, in a single market, during a period that included one major football international break and the run-up to the Advent season, when Croatian online traffic typically climbs. Replication matters. A hold that works at 3 seconds may do nothing at 2, and may actively annoy at 5. The dose-response curve is unmeasured, and the test only tells us that 3 seconds beat 0 seconds.

What a Croatian operator should take from this

The practical reading is narrower than the headline. A 3-second hold on withdrawal confirmation is a cheap, low-risk intervention that appears to reduce player-initiated cancellation of withdrawal requests by a meaningful margin, with the strongest effect on young players, small amounts, and weekend evenings. It is not a substitute for a cool-off period, and it should not be sold internally as one. It is a commitment device, and commitment devices work best on the players least equipped to commit.

The implementation details that separated the operators with the cleanest results were unglamorous. A visible fill animation with a distinct "keep holding" state. A haptic or visual confirmation at completion. A cancel path that was not hidden, moved, or made slower. And a label under the button explaining what the hold was for. The operators that skipped the label saw twice the support-ticket uplift.

The measurement discipline matters as much as the design. Report relative and absolute lifts together. Segment by amount and by time of day, because the average hides the effect. And instrument the initiation step, not just the completion step, or the selection effect will quietly inflate your result.

Responsible gambling teams should read the 19% as an invitation rather than a result. If a three-second press moves completion by 19%, the natural next question is what a three-second pause at the deposit confirmation does — and whether the same players who benefit from friction on the way out would benefit from it on the way in. Nobody in this test ran that arm. The operators involved have the data to run it now, and the fact that none of them has published a deposit-side result is itself worth noting.

The open question is not whether the hold works. It is what else in the cashier flow is quietly doing the same job, unmeasured, and whether the industry's current approach — treating withdrawal friction as a compliance overhead to be minimised — has been leaving a measurable number on the table for years.