A 0.9-second approval still beats a 12-hour manual review

Most regulated sportsbooks now settle in-play bets in under a second, and the fastest publicly benchmarked operators clear KYC checks in roughly 0.9 seconds. A manual review at the same operator, triggered when the automated system flags something, takes between 8 and 14 hours. That gap is the whole story: the automation is not a convenience feature, it's the product, and the human reviewer sitting behind it is now the exception path nobody wants to hit.

The two-second rule nobody wrote down

Somewhere between 2015 and 2020, bettors stopped tolerating latency. It wasn't a decision anyone announced. It just became true that if your cashout took 40 minutes, people opened a second account at a competitor and you never saw them again.

The numbers back this up more cleanly than the marketing does. In a 2023 sample of 2.4 million withdrawal requests across four European-licensed operators, the median automated payout landed at 4 minutes 12 seconds from request to confirmation. Manual review cases in the same dataset had a median of 11 hours 40 minutes. Same operators, same payment rails, same currencies. The only variable was whether a human touched the request.

That's not a modest difference. It's two different products wearing the same logo.

Why the automated path is fast

Automated settlement is fast because it answers a narrower question than people assume. It isn't asking "is this bettor trustworthy?" It's asking "does this transaction match the pattern of the last 200 transactions from accounts with this profile?" That's a lookup, and lookups are cheap.

Risk scoring at the point of withdrawal typically runs 40 to 60 checks in under 300 milliseconds: device fingerprint consistency, deposit-to-withdrawal ratio, payment method match, IP geolocation drift, velocity across the last 24 hours. Any single failure doesn't trigger a block. It adds points. Cross a threshold and the request falls out of the automated lane entirely.

The threshold is where operators quietly compete. Set it too tight and you review 6% of withdrawals manually, which is unsustainable at scale. Set it too loose and you eat fraud losses that show up two quarters later as a line item nobody can explain. Most mid-size operators land somewhere between 0.8% and 2.5% manual review rates. The best-run books I've seen data from sit under 1%.

What the 12-hour review actually is

Here's the part the industry doesn't say out loud: the manual review queue is not staffed for speed. It's staffed for cost.

A payments analyst working a review queue at a typical operator handles somewhere between 60 and 120 cases per shift. If you're running 1% manual review on 30,000 daily withdrawals, that's 300 cases a day, which means three to five full-time analysts just to keep the queue from growing. Add weekend coverage, language coverage across markets, and the overnight shift, and you're looking at a team of ten to fifteen people to process 1% of your traffic.

That team is expensive, and it's the first place a cost-cutting exercise looks. So you get understaffed queues, and understaffed queues produce the 12-hour median. Not because anyone decided 12 hours was acceptable, but because nobody decided it wasn't.

The compounding problem

Slow reviews don't just annoy the customer waiting. They poison the data the automated system learns from.

Every case that sits in the queue for half a day is a case where the customer's behavior after the flag is unobserved. Did they try again? Did they contact support? Did they churn? If you don't capture that, your threshold tuning is guesswork. And if your threshold tuning is guesswork, you flag too many people, which lengthens the queue, which makes the data worse.

I've watched two operators try to fix this by simply hiring more analysts. It works for about six weeks. Then the flag rate creeps up because the model is now trained on a larger volume of stale cases, and you're back where you started with a bigger payroll.

The 0.9-second claim, examined

The 0.9-second figure comes from a specific, narrow benchmark: the time between a bettor submitting a withdrawal and the operator's system returning an approval decision, measured at the API layer, excluding blockchain confirmation or bank settlement. It's an internal system metric, not a "money in your account" metric.

That distinction matters, and it's where a lot of operator marketing gets slippery. A 0.9-second approval decision followed by a 3-hour payment processor delay is still a 3-hour experience for the customer. The approval is fast. The money isn't always.

Still, the approval decision is the part the operator controls. Payment rails are mostly rented. So when an operator says "instant withdrawals," they usually mean "instant approval," and the honest version of that claim is worth measuring separately from the end-to-end time.

If you're evaluating an operator, ask for two numbers, not one: median approval time and median end-to-end settlement time. The gap between them tells you how much of the speed is theirs and how much is the payment provider's.

Where the 0.9 seconds breaks down

It breaks down in three predictable places:

• First withdrawal on an account. No history, no pattern to match against, so the score has nothing to work with. Expect manual review.
• Payment method mismatch. Deposited with card, withdrawing to crypto, or vice versa. This is a legitimate fraud signal and it costs you time.
• Geolocation drift. VPN use, travel, or a shared IP. Not necessarily suspicious, but it's a points-adder.

None of these are unreasonable triggers. The problem is that a bettor who hits one of them has no way to know they've hit it, and no way to resolve it except waiting. That's the part that needs fixing, and almost nobody has fixed it.

What would actually close the gap

The obvious answer is "make the manual review faster," and a few operators have tried. The results are mixed.

One approach that seems to work: route flagged cases to a tiered queue where low-risk flags get an automated decision with a 30-minute hold instead of a full human review. The hold gives the bettor time to respond to a prompt — "confirm this withdrawal method" — and if they do, the case clears automatically. If they don't, it goes to a human. In one operator's data, this cut manual review volume by 41% without increasing fraud losses over a six-month window.

The other approach, which is less popular because it's harder, is to fix the upstream data so fewer cases get flagged in the first place. That means better device fingerprinting, better payment method linking, and — critically — a feedback loop from resolved cases back into the scoring model. Most operators have the first two. Very few have the third.

The uncomfortable question is whether the 12-hour manual review is even a bug, or whether it's the cost of doing business in a regulatory environment that treats speed and fraud prevention as opposing goals. If a regulator audits you and finds you approved a flagged withdrawal in 90 seconds, that's a finding. If they find you took 12 hours, that's due diligence.

So the industry has settled on a compromise: automate the easy 98%, and let the hard 2% rot in a queue that nobody wants to fund. The 0.9-second approval is real. It's just real for the people who don't need it most.