Withdrawal Requests Drop 27% When ETA Copy Names the Bank

A payment operations team at a mid-sized European operator ran an A/B test in Q1 2025 across 41,000 withdrawal requests from Croatian, Slovenian and Austrian players. Half the requests went through a cashier flow that said "Withdrawal processing: 24–48 hours." The other half said "Withdrawal to your Zagrebačka banka account: typically arrives by Thursday, 14:00." The second group generated 27.4% fewer support tickets tagged "where is my money," and — more interesting for anyone running retention — 11.2% fewer voluntary reversals before the payout cleared. The wording of an ETA is not cosmetic. It is a risk control, a cost control, and arguably a compliance control, and most operators still treat it as a line of UX copy.

The mechanism behind that number is worth unpacking before we get to what to do about it. Croatian players in particular have a specific relationship with withdrawal timing: they are used to bank transfers that clear next business day, they are used to Aircash and Revolut behaving near-instantly, and they are used to a banking sector that has spent two decades tightening its consumer-facing SLAs. When an iGaming cashier presents a vague ETA, the player does not lower their expectation to match it. They import their expectation from every other financial transaction they make, and then they call support when reality diverges. Naming the bank does something subtler than setting a deadline: it tells the player that the operator knows where the money is going, which reframes the wait from "the casino is stalling" to "the bank is processing."

Why "24–48 hours" is the worst ETA you can print

Vague time ranges read as evasive, and players have learned to interpret them that way. A 2023 survey of 2,100 Croatian online players conducted by a Zagreb market research firm found that 63% believed a stated "24–48 hours" meant the operator would take the full 48 and possibly longer. Only 9% believed it meant 24. That gap between printed copy and player interpretation is where support volume lives.

There is also a factual problem. "24–48 hours" almost never describes the actual critical path. For a card withdrawal to a Croatian bank, the operator-side approval might take 40 minutes, the PSP batching might add 3 hours, and the acquiring bank's settlement cycle might add 1–2 business days. The number that matters to the player is the last one, and it is the one most cashiers refuse to name. When you print a range that spans the operator's own processing window and the bank's, you are averaging two unrelated clocks and producing a number that is wrong for everyone.

The A/B test data above is consistent with a broader pattern. In a 2024 review of 12 operators running similar tests, the average reduction in "where is my money" tickets when the ETA named the receiving institution was 19–31%, with the effect strongest on first withdrawals and weakest on players who had already withdrawn more than five times. That last detail matters: experienced players have calibrated their own expectations and no longer read the copy. New players do, and new players are the ones whose first withdrawal experience determines whether there is a second deposit.

The reversal problem is bigger than the ticket problem

Voluntary reversal — the player cancelling a pending withdrawal and returning the funds to their balance — is the single most expensive player behaviour in the cashier flow. It is expensive for the player, who usually loses the money. It is expensive for the operator, because a reversed withdrawal frequently becomes a complaint, a chargeback risk, or a churned account. And it is directly sensitive to ETA copy.

The 11.2% reduction in reversals from the A/B test is not explained by tickets alone. What appears to happen is that a specific ETA ("by Thursday, 14:00") creates a mental commitment point. The player has told themselves the money arrives Thursday. Reversing on Tuesday means the money is back in the balance, and the temptation to stake it is immediate. But the specific ETA makes the reversal feel like it breaks a plan rather than merely delaying an abstraction. Operators who have run reversal-reduction experiments consistently find that specificity outperforms urgency messaging, bonus offers to cancel the reversal, and even removing the reversal button entirely (which tends to push players to support, where the reversal is often granted anyway).

What "naming the bank" actually changes

The title of this piece says the ETA names the bank. It is worth being precise about what that means, because there are three distinct versions and they perform differently.

Version 1 — Institution name only. "Withdrawal to Zagrebačka banka." This is the weakest version. It confirms the operator knows the destination, but it does not commit to a time.

Version 2 — Institution plus day. "Withdrawal to Zagrebačka banka: arrives Thursday." This is where the measurable effect starts. The player now has a specific day to anchor on, and support tickets drop sharply because the player has a defined point at which to become concerned rather than an open-ended anxiety window.

Version 3 — Institution, day, and time window. "Withdrawal to Zagrebačka banka: typically arrives Thursday between 12:00 and 16:00." This is the version that produced the 27.4% ticket reduction. It is also the version that carries the most operational risk, because if you print a window you have to hit it, and hitting it requires actual data from your PSP about settlement times by receiving bank.

That last point is the reason most operators stop at Version 1 or 2. Version 3 requires you to know, with reasonable confidence, when a payment to a specific Croatian bank will land. That knowledge exists — PSPs have it, and most can expose it via API or a daily settlement report — but it is rarely plumbed into the cashier. The operators getting the 27% number are the ones who did that plumbing.

The compliance angle nobody talks about

Under Croatian gambling regulation and the broader EU AML framework, operators are required to process withdrawals in a timely manner and to be transparent about processing times. The current standard practice — a generic range — is technically compliant. But the direction of travel in regulatory guidance across the EU is toward specificity: the player should be able to understand when and how their money will arrive. An operator that can demonstrate it names the receiving institution and provides a realistic window is in a stronger position in any dispute than one that printed "24–48 hours" and took 71.

There is also a responsible gambling dimension that is easy to miss. A player waiting on a withdrawal is a player not currently staking. The longer the wait, the more likely they are to reverse and re-stake, which is exactly the behaviour that problem gambling screening is designed to catch. A clear, specific ETA reduces the ambiguity window in which reversal is tempting. That is not a marketing argument. It is an argument that the cashier copy is part of the safer gambling toolkit, and should be reviewed by whoever owns that function, not only by whoever owns conversion.

How to build an ETA that names the bank

The operational recipe is less exotic than it sounds, but it requires three data sources to be joined.

First, the receiving institution. You already have this. When a player enters IBAN HR[digits], the first five digits after the country code identify the bank. Croatian IBANs are structured so that the bank code is deterministic. You do not need to ask the player which bank they use; you can derive it. The same is true for most EU IBANs. For card withdrawals, the issuing bank is available from the BIN, though it is less reliable for naming purposes because players rarely recognise their card issuer by the BIN's institution name.

Second, the settlement profile for that institution. This comes from your PSP. Most PSPs maintain, or can generate, a per-receiving-bank settlement profile showing median and 90th-percentile arrival times. For Croatian banks, the spread is meaningful: some institutions settle same-day if the payout is submitted before a cutoff, others batch overnight, and a small number of smaller banks and credit unions take an extra day. If your PSP cannot provide this, ask for the raw settlement reports and build it yourself. Two weeks of data is enough to produce a usable median.

Third, your own processing cutoff. This is the time by which a withdrawal must be approved to make that day's settlement run. It is usually earlier than operators assume, and it is often the single largest source of delay. A withdrawal approved at 15:05 that misses a 15:00 cutoff does not arrive a few minutes late; it arrives a full business day late. Printing the cutoff in the cashier — "requests approved before 15:00 arrive Thursday" — is one of the highest-leverage pieces of copy available, and almost nobody does it.

The failure modes to avoid

Naming the bank introduces new ways to be wrong, and being wrong is worse than being vague. Three failure modes are common.

The stale ETA. Settlement profiles change. A bank changes its batch window, a PSP changes its cutoff, a public holiday intervenes. If your ETA copy is hardcoded, it will drift out of accuracy within a quarter. The operators doing this well regenerate the copy from data on a schedule — weekly for the day-of-week component, daily for the cutoff — rather than treating it as a static string.

The over-promise. Printing the median arrival time means half your players wait longer than promised. Print the 75th or 90th percentile instead, and consider showing a range that reflects the real distribution. A player told "Thursday" who receives the money Wednesday is delighted. A player told "Thursday" who receives it Friday is on the phone. Asymmetry of disappointment is real, and it argues for conservative ETAs.

The wrong bank name. Croatian players know their banks. If you print "Zagrebačka banka" for a player who banks with Erste, you have not built trust; you have demonstrated that your system is guessing. Derive the name from the IBAN, not from a lookup table keyed on something less reliable, and if the derivation fails, fall back to a generic ETA rather than a wrong specific one.

What the 27% is actually worth

It is worth translating the ticket reduction into money, because the case for doing this work is usually made on conversion and it should be made on cost.

A mid-sized operator processing 41,000 withdrawals per quarter, as in the test above, might see 6–9% of those generate a support contact under the vague-ETA condition. That is roughly 2,500–3,700 tickets per quarter. At a fully loaded cost of €4–6 per ticket for chat and email handling, the vague ETA is costing €10,000–22,000 per quarter in support alone. The 27.4% reduction recovers €3,000–6,000 of that. That is not a transformative number on its own.

The reversal reduction is where the real money sits, and it is harder to quantify because it depends on the average reversal amount and the downstream behaviour. If 11.2% fewer reversals translates to, say, 400 fewer reversals per quarter at an average of €180, that is €72,000 in funds that stayed in the withdrawal pipeline rather than returning to the balance. Some of those players would have withdrawn again anyway. Some would have lost the reversed funds and churned. The honest estimate is that the operator retains somewhere between €15,000 and €40,000 per quarter in value that would otherwise have leaked, plus a meaningful reduction in complaint-handling load and a modest improvement in first-withdrawal-to-second-deposit conversion.

The point is not that this is a huge win. It is that it is a cheap win. The work is a data join, a copy change, and a monitoring loop. There is no new licence, no new vendor, no new game. For an operator already struggling with cashier support costs and reversal rates, it is one of the better returns available on a product team's time.

The Croatian specifics that matter

Croatia has a few characteristics that make this more valuable here than in some other markets. The banking sector is unusually concentrated, which means the number of settlement profiles you need to maintain is small — a dozen institutions cover the large majority of players. The IBAN structure is clean and deterministic. And Croatian players have high expectations of payment speed because domestic instant payment infrastructure has improved markedly over the past five years, which means a vague ETA is compared against a fast baseline rather than a slow one.

The countervailing factor is that Croatian players are also relatively experienced with the difference between operator processing time and bank settlement time, particularly those who have withdrawn from more than one operator. That experience cuts both ways. It makes them less likely to raise a ticket over a one-day delay, but it also makes them more likely to reverse if they suspect the operator is stalling. Specific copy addresses both.

Where this leaves the cashier

The 27% figure is a useful headline, but the underlying finding is narrower and more durable: players respond to information about their own money, and they respond badly to the absence of it. Naming the bank is one way to supply that information. Naming the cutoff time is another. Showing a per-institution arrival window is a third. None of them are technically difficult, and all of them are cheaper than the support tickets they prevent.

The open question is whether the effect holds as players learn that the ETA is data-derived rather than a promise. If operators start printing specific ETAs and then miss them — because the data is stale, or the cutoff moved, or the PSP changed its batching — the specificity stops being reassuring and starts being evidence of incompetence. The 27% reduction is real, but it is conditional on the ETA being right. The operators who treat this as a copy change will get a quarter of benefit and then watch it decay. The ones who treat it as a data product — with owners, refresh cadence, and accuracy monitoring — will keep it.