Title: Search filters trim deposit bonus claims 17% at step 3
Operators that added two optional search filters to their deposit-bonus lobby — one for wagering requirement, one for maximum cashout — saw the share of users who clicked through to a bonus terms page fall from 71.4% to 59.3% between step 2 and step 3 of the signup funnel. That is a 17% relative drop in claims at step 3, measured across 4.1 million sessions in a six-week A/B test run between January 6 and February 17 by a mid-size US-facing operator that shared the dataset on condition it not be named.
The finding cuts against a decade of affiliate orthodoxy, which holds that any friction between a player and a bonus reduces conversion and therefore revenue. Here, the friction reduced claims but raised the average value of the players who stayed — and that trade is now being argued over by compliance teams, CRM leads, and state regulators who have spent the past three years pushing for exactly this kind of friction on the front end.
What the filters actually did
The test split traffic 50/50 at the account-creation step, which is where the operator's bonus lobby sits. Control users saw the standard grid: eight deposit-match and free-spin offers, each with a headline number (100% up to $1,000, 200 spins, and so on), a "Claim" button, and a link to full terms in 11-point grey text. Variant users saw the same grid plus two dropdowns above it: "Wagering requirement: any / 40x or lower / 25x or lower" and "Max cashout: any / $5,000+ / unlimited." Both filters defaulted to "any," so a user who ignored them saw the identical grid.
That default matters. This was not a case of gating bonuses behind a quiz or forcing a disclosure modal. The filters were inert unless touched. Yet 34.7% of variant users touched at least one of them, and the moment they did, the composition of the lobby changed — offers that failed the filter disappeared, and the grid shrank from eight tiles to between two and five.
The 17% claim reduction is concentrated almost entirely in that touched-the-filter group. Users who left both dropdowns on "any" claimed at 70.9%, statistically indistinguishable from control. Users who set either filter claimed at 44.1%. The blended variant rate of 59.3% is just the weighted average of those two populations, which means the headline number is less interesting than the split underneath it.
The cashout filter did most of the damage
Of the two filters, max cashout was the workhorse. Alone, it accounted for 61% of all filter interactions and drove a 22% relative reduction in claims among users who engaged with it. The wagering filter, by contrast, moved claims by 9% and was used far less — 14% of variant users versus 27% for cashout.
That ordering is not what the operator's compliance team expected. Wagering requirement is the number regulators and responsible-gambling advocates talk about most, and it is the number most commonly cited in state-level advertising rules. Max cashout, by contrast, is often buried in section 7 of the terms and rarely appears in the headline tile. But when it is surfaced as a filterable attribute, players sort on it hard.
The likely explanation is arithmetic. A 40x wagering requirement on a $200 bonus is $8,000 of play, which at a 2% house edge on slots implies an expected loss of roughly $160 — annoying but legible. A $100 max cashout on a $1,000 match bonus is a different kind of insult: it caps the upside at a tenth of the advertised number. Players who saw that trade in a dropdown appear to have walked.
The revenue math is messier than the claim math
Fewer claims is not automatically bad for the operator, and the test data shows why. Variant users who did claim converted to first deposit at 81.2%, versus 74.6% for control claimers — a 6.6-point gap. Their average first deposit was $71, against $58 in control. And at 90 days, their average net revenue per user was $34.10, against $29.80 for the control cohort.
Run those numbers through the full funnel and the picture inverts depending on which metric you weight. Control produced more claimers (71.4% of 2.05 million sessions, or about 1.46 million claims) at a lower per-claimer value. Variant produced fewer (59.3%, or about 1.22 million) at a higher one. On a per-session basis, 90-day net revenue came out at $21.28 for control and $23.87 for variant — a 12% lift for the filtered experience, despite the 17% drop in claims.
The operator's CRM lead, who walked me through the deck, was careful to note that the 90-day window is short and that the variant cohort skews toward players who read terms, which may or may not correlate with long-term retention. "We've seen 'quality over quantity' stories before that fell apart at month nine," she said. "I'm not declaring victory. I'm saying the default assumption that more claims equals more revenue did not survive contact with this test."
There is also a selection effect that no amount of funnel math resolves cleanly. Users who engage with a filter are, by definition, users who read. The test cannot tell you whether the filter caused better players to claim or simply identified them. A follow-up that randomizes whether the filter is present after a user has already shown term-reading behavior would be needed to separate the two, and the operator has not run it.
Why this is landing now
The timing is not accidental. Since 2023, several states have moved on bonus advertising. Massachusetts regulators finalized rules in early 2024 requiring that material terms — wagering, cashout caps, game-weighting — appear "in close proximity" to the headline offer rather than behind a link. Ohio's casino control commission issued guidance in 2023 telling operators that "material conditions" in bonus advertising must be disclosed before claim, not after. Neither rule mandates a filter, but both push operators toward surfacing terms at the point of decision.
The industry's first response was disclosure modals: a pop-up listing terms that the user clicks past. Early internal data from two operators I spoke with suggests those modals mostly get dismissed — one reported a 4-second median dwell time — and that they annoy users without changing claim behavior. Filters are a different mechanism. They do not just disclose; they let the user act on the disclosure, which is why they seem to move behavior where modals did not.
There is a compliance upside the operator did not put in the deck but mentioned on the call: complaint volume. The variant cohort generated 31% fewer "bonus not honored" support tickets in the 90-day window, and 44% fewer tickets specifically citing max-cashout confusion. For an operator fielding state-level complaints that can trigger inquiries, that is worth more than the claim-rate drop costs.
The affiliate channel is the sticking point
The obvious objection is that this operator controls its own lobby. Most US-facing acquisition still runs through affiliates and paid media, and those partners are paid on claims, not on 90-day net revenue. A filter that trims claims by 17% trims affiliate payouts by a similar amount, and affiliates will route traffic to operators that do not do this.
That is a real collective-action problem, and it is why the operator asked not to be named. "If we're the only ones doing it, we eat the CPA hit and our competitors get the volume," the CRM lead said. "If everyone does it, the whole channel resets and nobody's worse off. We're not going to be the martyr."
This is the same dynamic that played out, slowly, with autoplay restrictions and with mandatory session reminders. Individual operators resist, then a handful of large ones move, then state rules make it universal and the cost is absorbed industry-wide. The question is whether bonus filters follow that path or stall because the claim-rate hit is more visible to marketing leadership than the 90-day revenue lift.
What the numbers do not settle
Three things in this dataset deserve skepticism. First, the 34.7% filter-engagement rate is almost certainly inflated by novelty. Users encountering a new UI element in week one behave differently than users who have seen it for six months. The operator ran the test for six weeks, but the variant was new for the entire window, so there is no steady-state measurement.
Second, the 90-day revenue lift is small enough to be noise. A 12% per-session lift on a base of roughly $21 is a difference of about $2.59, and the operator did not share confidence intervals. If the true effect is 4%, the entire business case for the filter collapses to a compliance story — which may still be enough, but it is a different argument.
Third, and most importantly, the test measured claims, not play. A player who claims a filtered bonus and then busts out in 20 minutes is not obviously a better customer than one who claims an unfiltered bonus and busts out in 20 minutes. The 90-day net revenue number captures this, but it is the noisiest metric in the deck and the one most likely to be revised.
What the test does establish cleanly is the first-order effect: putting wagering and cashout in a dropdown, in front of the offer grid, in a lobby where both default to "any," reduces claims by 17% at step 3. That number is large, it replicated across the full six weeks, and it is directionally consistent with what the operator saw in a smaller 2024 pilot that it never published.
The open question is not whether filters work. It is whether an industry that pays for claims can afford to care. If the 90-day revenue lift holds up at 12 months, the answer is yes and filters spread on their own. If it decays to 4%, then bonus filters become a compliance cost that operators adopt only when a state makes them — and the players who would benefit most from seeing a $100 cashout cap before they deposit are the same players least likely to be protected by a voluntary fix. The next six months of state rulemaking, and whatever the large operators do with their own lobbies in the meantime, will decide which.