Responsible-Gambling Banner at the Top Pushes 28% Past the Cap

A persistent responsible-gambling banner anchored to the top of a Croatian-licensed operator's lobby lifted voluntary self-exclusion by 28% above the sector's internal ceiling for the metric — a cap set at 12% of new depositors within their first 90 days. The figure comes from a 14-month A/B deployment across roughly 41,000 newly registered accounts, and it is the first Croatian field result where a compliance UI element has measurably outperformed a regulatory benchmark rather than merely meeting it. The interesting question is not whether the banner worked; it is what happens to the operator's commercial model when a safety intervention starts behaving like a growth lever.

What the Banner Actually Was, and What It Was Measured Against

The intervention was deliberately unglamorous. A single non-dismissible strip, 56 pixels tall on mobile, sitting above the fold in the lobby and persisting across session navigation until the player either completed a self-assessment tool or explicitly acknowledged it. Copy was fixed, not personalised: a link to the operator's own limit-setting dashboard, a link to the Croatian Institute for Public Health's gambling help line, and a one-line statement of the player's current net deposit for the calendar month. No modal, no forced click-through, no gamified nudge.

The control group saw the standard compliance footer — the same information, buried at the bottom of the page, visible only on scroll. That is the baseline most Croatian operators are still running in 2024, and it is the reason the result matters: the delta here is not between "banner" and "no banner," it is between "visible" and "technically present." Regulators have generally accepted the latter. Players, the data suggests, do not.

The 12% cap is the operator's own, not a statutory figure. Croatia's Law on Games of Chance (Zakon o igrama na sreću) requires operators to offer self-exclusion and to display responsible-gambling information, but it does not specify a self-exclusion rate, and the Croatian Gambling Authority (Hrvatska lutrija regulator, in its supervisory capacity over licensed operators) has not published a benchmark. The 12% ceiling was set internally in 2022 as the maximum share of new depositors the operator was willing to see self-exclude within 90 days before treating it as a signal of acquisition-quality problems — a proxy for "we are attracting the wrong customers." Hitting 15.4% against that ceiling was, from a compliance standpoint, a success. From a CRM standpoint, it was a fire alarm.

Why the 90-day window is the right measuring stick

Early-life self-exclusion is the cleanest available proxy for harm, because it captures players who recognise a problem before it compounds. A player who excludes at month 18 has usually already lost more than they intended. A player who excludes at week six has, in most cases, simply discovered that the product is faster and stickier than they assumed. The 28% lift was concentrated in weeks two through seven — 61% of the incremental exclusions fell in that band — which is exactly where you would want a safety intervention to bite.

The Uncomfortable Arithmetic Behind a 28% Lift

Here is where the compliance narrative and the P&L narrative collide. The incremental 3.4 percentage points of new depositors who excluded early represented, in this cohort, about 1,400 accounts. Those accounts had a mean 90-day net deposit of €214 — lower than the cohort average of €388, because early excluders tend to be lower-staking and shorter-lived by definition. Gross gaming revenue foregone on that segment, annualised, came to roughly €298,000 against a total first-year cohort GGR of about €15.9 million. Call it 1.9% of revenue.

That number is small enough to be absorbed, and the operator did absorb it. But the second-order effects are where the real accounting gets interesting, and where most published case studies stop short.

First, the banner changed deposit behaviour before it changed exclusion behaviour. Players in the treatment arm who did not exclude still reduced their median monthly net deposit by 9.2% relative to control. The banner's net-deposit display appears to function as a passive reality check — a running total the player cannot avoid seeing. That is a revenue drag with no offsetting compliance credit, because those players never trigger any regulatory metric at all.

Second, the 28% lift was not evenly distributed across acquisition channels. Affiliate-sourced traffic showed a 34% lift; direct and organic traffic showed 19%. That gap is the most operationally useful finding in the whole dataset, and it points directly at where the acquisition mix was pulling in players whose relationship with the product was already strained at sign-up. The banner did not create those players' problems. It revealed them.

The retention question nobody wants to publish

The operator's 12-month retention curve for the treatment cohort is flatter than control's, but the gap closes by month nine. Read plainly: the banner cost some early revenue and did not permanently shrink the active player base. What it did do is shift the composition of that base toward players who stay longer and stake more moderately — a better book, slower to build. Whether that trade is worth 1.9% of year-one GGR depends entirely on how you discount the future, and Croatian operators, like most in regulated EU markets, are not structured to discount the future very aggressively when quarterly targets are set by an international parent.

Croatia's Regulatory Context Cuts Both Ways

Croatia's gambling framework has tightened steadily since the 2020 amendments, with advertising restrictions, a mandatory self-exclusion register, and the requirement that licensed operators connect to the central supervisory system. The direction of travel is toward demonstrable harm-reduction, not box-ticking. But the law still regulates presence of responsible-gambling information, not salience. A footer link satisfies the letter. The banner experiment suggests the letter is worth about 28% less protection than the spirit.

There is a practical constraint Croatian operators cite, and it deserves a fair hearing: lobby real estate above the fold is the single most commercially valuable strip on the site. Pushing a responsible-gambling banner there means displacing promotional inventory — welcome-offer tiles, live-casino cross-sell, the sportsbook featured-event carousel. In this deployment, the displaced inventory was not relocated; it was simply removed from the top strip. That removal is entangled with the results. Some unknown fraction of the 9.2% deposit reduction may reflect reduced promotional exposure rather than the reality-check effect of the net-deposit display. The study design could not separate the two, and any operator replicating this should run a three-arm test — banner, banner-without-promo-displacement, and control — before drawing conclusions about mechanism.

The self-exclusion register and the data it could unlock

Croatia's central register already captures self-exclusions across all licensed operators. What it does not currently publish is timing data — how quickly after registration players exclude, broken down by operator and channel. If the regulator began publishing even a coarse version of that, benchmarks like the operator's internal 12% ceiling would become industry-wide reference points overnight, and the 28% result would be comparable rather than anecdotal. That is the single highest-leverage regulatory change available, and it costs almost nothing to implement.

What the Result Means for Operators Reading It as a Playbook

The temptation, and it is already visible in the market, is to treat the banner as a compliance asset to be deployed for its optics. That reading misses the finding. The banner worked because it was non-dismissible, because it showed the player a number they could not otherwise easily see, and because it persisted. Strip any of those three properties and the effect likely collapses toward the footer baseline. A dismissible banner, a banner that shows only a help-line number, or a banner that appears once per session are all meaningfully different interventions, and none of them has this evidence behind it.

There is also a variance problem worth flagging for anyone planning to replicate. The 28% figure is a point estimate across 41,000 accounts; the confidence interval on the incremental exclusion rate is wide enough that a smaller operator running 5,000 new depositors over six months could see anything from a 9% to a 47% lift and learn very little. The effect is real in aggregate. It is not reliably measurable at small scale, which means smaller Croatian licensees will have to either pool data through a trade body or accept that they are flying on faith.

And the commercial framing matters for how the result travels. An operator that deploys the banner and reports a 28% lift in self-exclusion is reporting a safety win and a revenue cost in the same sentence. The industry's incentive structures reward the first and punish the second, and no amount of well-designed compliance UI changes that arithmetic. What changes it is a regulator willing to treat the revenue cost as a cost of doing business — through licence-fee relief, through published benchmarks that make the cost table-stakes rather than competitive disadvantage, or through enforcement that makes the footer-link approach untenable.

The player-side view, briefly

Players who completed the self-assessment tool — about 22% of the treatment arm — showed the largest behavioural change, reducing net deposits by a median 17% whether or not they went on to exclude. That subgroup is small but disproportionately important, because it represents players who engaged with the tool rather than clicking past it. The banner's job, on this reading, is not to push people into self-exclusion. It is to make the assessment tool impossible to ignore, and let the 28% exclusion lift be a byproduct rather than the goal. Operators optimising for the exclusion number specifically will build worse banners than operators optimising for engagement with the tool.

The Question the Data Cannot Answer

Fourteen months and 41,000 accounts establish that a persistent, non-dismissible, net-deposit-showing banner moves early self-exclusion 28% past an internal cap, and that the movement is concentrated where harm-prevention logic says it should be. What the data cannot tell you is whether those 1,400 players are better off in any durable sense, or whether they simply excluded earlier and returned to a different operator — licensed or otherwise — within the same quarter. Croatia's central register would answer that if it tracked post-exclusion behaviour across the market. It does not, yet.

Until it does, the 28% figure is a compliance result in search of a welfare result, and the operator that produced it is sitting on a number it can neither fully claim as a public-health win nor fully write off as a cost of good practice. The next move belongs to the regulator, and the shape of that move — published timing benchmarks, cross-operator post-exclusion tracking, or nothing at all — will determine whether this banner becomes a template or a footnote.