Provider Logos Load Before RTP Text, 35% Pick the Wrong Table

Provider branding is doing more work on Croatian-facing casino lobbies than the information players actually need to choose a game. In a session-level test of 40 lobby and table layouts across 12 sites licensed by the Croatian regulator, the average time from page paint to a fully rendered provider logo was 0.31 seconds, while the RTP figure for the same game appeared at 1.94 seconds — or not at all on 9 of the 40 layouts. When 35% of testers shown a two-table grid chose the lower-RTP option after the logos had loaded but before the return figures did, the ordering stopped looking like a design quirk and started looking like a cause.

That 35% figure comes from a small sample — 120 Croatian-resident adults with prior casino experience, recruited in Zagreb and Split in February 2024, each shown four randomised two-table grids with the same game type, same stake range, and RTPs of 96.4% and 94.1%. The study is not peer-reviewed and the sample is too thin to generalise to the whole market. It is, however, consistent with what the lobby data shows about load order, and load order is something operators control.

What the load-order data actually shows

The 40 layouts broke down into three rough groups. Sixteen rendered the provider mark inside the first viewport before any text about the game; fourteen rendered logo and RTP text together in the same paint; ten rendered RTP text first or in a fixed panel that did not move as the page settled.

The timing gap between logo and RTP text matters more than it sounds. In the sixteen logo-first layouts, the median gap was 1.6 seconds. That is roughly the time a thumb takes to travel from a screen edge to a tile on a 6.1-inch phone held one-handed — which is the posture most Croatian mobile players are in when they open a lobby on a tram or between tasks. The logo arrives while the finger is still moving; the RTP text arrives after the tap in many cases.

Three of the twelve operators placed RTP in a hover state on desktop, which means it does not exist at all on touch. Two placed it in a collapsed "game info" accordion below the fold. One — a mid-sized operator with a Croatian licence and a Serbian-facing sister brand — displayed "RTP 96.1%" in the tile but served a demo build at 94.8% on the same URL until a session cookie was set. That last case is worth separating from the others: it is not a load-order problem but a disclosure problem, and it falls under different obligations.

The provider logo, by contrast, was above the fold on all 40 layouts, at an average rendered size of 84×28 pixels on mobile. It is the single most consistently visible piece of information in a Croatian online casino lobby after the game name itself.

Why logos load first

Provider marks are almost always served from a CDN as small SVG or WebP assets and cached across sessions, so they are cheap to paint. RTP figures are usually pulled from a game-metadata endpoint that sits behind the operator's own API, sometimes with a per-jurisdiction override layer. That layer is where Croatian RTP disclosure lives, and it is also where the latency lives.

There is a second reason, less technical. Provider logos are commercial assets supplied by the studios under partnership terms; operators have an interest in their prominence and studios audit for it. RTP text has no such sponsor. When a layout is assembled under deadline, the element with a contractual owner gets placed above the fold and the element without one gets pushed into the metadata panel. Nobody has to decide to bury the RTP; it just loses the argument.

The 35% result and what it does and does not prove

The two-table test was built to isolate ordering. Each participant saw four grids: two with logo-first rendering and a 1.5-second delay before RTP text appeared, and two with logo and RTP text rendering in the same frame. The game pairs were matched for theme, volatility band, and stake range, so the only difference within a pair was the return figure. Choices were logged by tap position, not self-report.

In the simultaneous-render grids, 22% chose the lower-RTP table. In the logo-first grids, 35% did. The 13-point gap is the finding. It is large enough to be interesting and small enough to be fragile: with 120 participants across four grids each, the confidence interval on a 13-point difference is wide, and a replication with 400 or more participants would be needed before anyone should cite the number in a regulatory submission.

Two caveats deserve stating plainly. First, the 22% baseline is not zero, which means a meaningful share of experienced players pick the lower-return table even when the information is right there — usually because of theme, volatility, or a bonus feature they recognise. Ordering is not the whole story and probably not the largest part of it. Second, the test used a 2.3-point RTP gap (96.4% vs 94.1%), which is at the wide end of what players actually encounter within a single game type. Narrow the gap to 0.4 points and the effect likely shrinks; widen it and the effect may not grow, because a large gap is easier to spot once the text lands.

What the test does support is narrower and more useful: the ordering of information changes choice distribution among players who have not yet read the return figure. That is a description of a window — roughly 1.5 to 2 seconds — during which the lobby is making an argument the player has not been given the counter-argument to.

The Croatian context

Croatia's online gambling market is regulated by the Ministry of Finance under the 2020 Law on Games of Chance, with licensed operators required to display game rules and return information. The law does not specify load order, viewport position, or minimum render time, and it is not obvious that it should — regulating paint sequence would be an odd instrument. But the gap between "displayed" and "displayed before the decision" is exactly the kind of gap that compliance checklists miss.

Croatian players also skew toward mobile more heavily than the EU average in the sessions I looked at: 71% of lobby sessions in the sample were on a phone, against a rough EU figure nearer 62%. Mobile is where load order bites hardest, because there is no hover state and no wide viewport to hold a side panel.

Where the responsibility sits

Three parties could change the ordering, and each has a reason not to be the first.

Operators control the layout. They also control the metadata layer and could, at modest engineering cost, ship RTP text as part of the initial tile render rather than as a follow-up fetch. The argument against is commercial: a tile with a 94.1% figure next to a 96.4% figure sells the second game less well than a tile with two logos. That argument is real and nobody in the industry says it out loud.

Studios supply the logos and, in most contracts, the RTP metadata. Some studios already publish RTP ranges per game and per market. A studio that shipped RTP as part of the logo asset — same request, same cache, same paint — would close the gap without asking operators to make a commercial choice.

The regulator could require that return information be present in the first rendered frame of any game tile, which is a testable and specific rule. It is also the kind of rule that is easy to write and hard to audit without automated tooling, and Croatian enforcement capacity is not obviously built for frame-level checks.

None of these is a complete fix. The 22% baseline in the simultaneous-render condition is a reminder that even perfect disclosure leaves a fifth of experienced players choosing on other grounds — which is their right. The target is not to make everyone choose the highest-RTP game. It is to stop the lobby from making the choice on the player's behalf in the 1.5 seconds before they have the number.

A note on responsible gambling

None of this is an argument that a player who picks a 94.1% game over a 96.4% one is being harmed in the way that problem gambling harms people. A 2.3-point RTP difference on a €50 session is a couple of euros in expected loss, and players are entitled to choose a game because they like the theme. The concern is cumulative and structural: if the lobby consistently front-loads the commercial signal and back-loads the factual one, the player's model of what they are doing is built on the wrong inputs. Croatian operators are required to offer self-exclusion and limit-setting through the national system, and those tools matter more than tile ordering — but tile ordering is the part that operates on every session, including the ones where nothing goes wrong.

What a fixable version looks like

The engineering fix is smaller than the argument around it. A tile that renders name, provider mark, and RTP in a single frame requires the RTP to be in the initial payload rather than a follow-up call. For a lobby of 200 tiles, that is a handful of extra kilobytes and one fewer round trip — in most cases it would make the lobby faster, not slower, because it removes a second request per tile.

The harder fix is cultural. Provider prominence is negotiated; RTP disclosure is complied with. As long as those two things sit in different conversations inside an operator, the logo will keep winning the race to the screen. The 35% figure is not proof of intent. It is proof that when nobody owns the ordering, the ordering defaults to the party with a commercial stake in it.

A reasonable next test would run the same grid with a 0.5-point RTP gap and 400 participants, and run it on a real lobby rather than a mock, with real tap logging. If the effect holds at that gap, the case for frame-level disclosure gets much stronger. If it collapses, then the 35% is mostly a story about obvious differences being easier to miss when they arrive late — still worth fixing, but a smaller claim.

The open question is whether Croatian operators will treat load order as a compliance surface before a regulator decides it is one. The tooling to check it exists and is cheap. The will to run it is the part that has not shipped.