Online Casino Profit Margins: what operators actually keep

Ask what margin an online casino makes and you get four answers, all technically true, because the industry calls four different numbers revenue. This page walks the money from a wager to retained profit. Every figure is cited to a regulator or an audited filing; where none existed, the line is given as a basis with no number.

What does "profit margin" mean in an online casino?

Turnover, GGR, NGR and EBITDA are four different numbers

Turnover is every stake placed, and the least informative: a player recycling $100 through twenty spins generates $2,000 of it. Gross gaming revenue is turnover less winnings, and what most regulators tax. Net gaming revenue is GGR less the direct costs of producing it. EBITDA is the only one that says whether the business makes money.

Which one a vendor quotes you, and why it matters

Platform revenue share is struck against NGR, but that definition is contractual and often negotiated in the supplier's favour. Three questions settle it: is duty deducted before or after the share, are bonus costs deductible and at grant or at wagering, and is there a monthly minimum?

How does money flow from a wager to retained profit?

The table is structural rather than benchmarked. Run it on your own inputs: the shape is the same for every operator, while each deduction's size is specific to you.

The waterfall is multiplicative, so a small deterioration early compounds through every later stage, and the early deductions are the least controllable: RTP is set by the game, duty by statute. Your levers sit lower down.

What hold rate should you expect by vertical?

Hold rate is gross gaming revenue divided by amount wagered. Few regulators publish it, because most publish revenue without the wagering behind it. Ontario publishes both, so every row below can be reproduced. Values are Canadian dollars, unaudited per the publisher.

Slots, live casino, table games, sportsbook

One limitation is load-bearing: Ontario reports slots, live dealer, RNG tables and peer-to-peer bingo as one "Casino" category. Anyone quoting separate published slots and live casino hold should be asked which regulator publishes it. The split is omitted, not estimated.

Volatility matters more than level. Over eighteen months casino hold stayed inside 0.37 percentage points while betting hold spanned 5.08 points, from 4.02% to 9.10%. Casino revenue can be forecast from wagering volume; sportsbook revenue cannot, because results move it.

Why RTP is not the same as your hold

A slot at 96% RTP has a 4% theoretical edge, but that is a long-run expectation for one game, not a portfolio hold rate. The larger trap is definitional: Ontario is explicit that "cash wagers do not include promotional wagers (bonuses)". Confirm any two hold numbers treat bonus wagering alike. Often they do not.

What comes out between GGR and NGR?

Bonus cost and the difference between granted and wagered

A bonus granted is not a bonus cost. The cost is what it pays out once wagering requirements are met, so model granted value for the accounting entry and wagered-through value for the economics. The tax treatment is what operators get wrong: HMRC is explicit that UK Remote Gaming Duty applies to freeplays, with tax due on "a customer's first use", so a bonus is dutiable even where it never converts.

Gaming tax and licence-jurisdiction effects

Duty is the largest single deduction from gross win in most regulated markets, and it changes. HMRC's published rates put Remote Gaming Duty at 15% before April 2019, 21% from that date and 40% on or after 1 April 2026. At 40%, gross win must grow by roughly a third just to hold net revenue flat.

Rates, bases and bonus deductibility vary by jurisdiction, and some markets tax stakes rather than gross win, penalising low-margin products disproportionately. Model each licence separately rather than applying a blended rate, and settle the licence jurisdiction and tax position before committing to a market.

Payment processing and chargebacks

Payment cost is a percentage of transaction value plus a fixed fee, and gambling merchants pay more because the schemes treat the category as high risk. No primary source for gaming-specific rates could be obtained, so none is quoted. The fixed component makes small deposits expensive, and a failed payment costs twice. See our guide to payment processing costs.

What are the operating costs below NGR?

No percentages appear there because none could be sourced to a public primary document. Published operator accounts show the aggregate shape instead.

Platform and content revenue share

These sit inside cost of revenue alongside gaming duty, which is why that line dominates a gaming income statement. Rush Street Interactive reported 65.4% of revenue in FY2025 against 68.5% in FY2020; DraftKings reported 58.7%, down from 62.5% in FY2023. It improves slowly and never approaches zero. For contrast Evolution AB, a supplier rather than an operator, reported a 65.9% adjusted EBITDA margin in Q2 2026.

Acquisition, and why CPA and payback dominate the model

Rush Street spent 14.5% of revenue on marketing in FY2025, down from 39.0% in FY2021; DraftKings spent 22.8%, down from 32.8% in FY2023. Heavy acquisition first, margin later. That is why payback period, not CPA alone, is the number to manage. Settle the affiliate commission models before signing. Rush Street reached a 7.7% operating margin in FY2025 after negative 47.9% in FY2020; DraftKings, on $6.05bn, still recorded negative 0.3%.

Which levers actually move the margin?

Retention over acquisition

Acquisition cost is paid once per player; every other deduction recurs. A player who stays twelve months rather than three produces roughly four times the gross win for the same acquisition cost. This is why the operators above improved margin by cutting marketing, not by raising hold. See our guide to player retention.

Bonus discipline and wagering requirements

Bonus cost is the most controllable large line and the one most often managed by instinct. Track granted value, wagered-through value and cost per retained player by campaign; offers identical on face value routinely differ by a multiple on realised cost. Where duty applies to freeplays, an ungoverned programme raises a duty liability too.

Payment mix and failure rates

Method mix changes both the fee and the conversion rate, and the second effect is usually larger: a lower headline rate with a worse authorisation rate costs more overall. Measure cost per successful deposit, not fee per transaction. Small-deposit cohorts can be unprofitable on payment economics alone.

How do margins differ by model and market?

White label, turnkey and owned-licence economics

Under a white label the licence, and often the payment relationships, belong to the provider, and its share of NGR reflects that: low entry cost, high ongoing cost. An owned build reverses it, so it wins only above a threshold you should calculate rather than assume: at what monthly NGR does the revenue share exceed the fixed cost of running it yourself? See our note on white label platform economics.

Regulated versus grey-market tax drag

The visible difference is duty: 40% of gross win in the UK against a materially lower rate elsewhere. The invisible ones matter more. Regulated markets bring compliance and reporting cost, and payment acceptance that grey markets do not. A lower tax rate on revenue you cannot bank is not a better margin.

How do you build a margin model before you launch?

1. Forecast turnover, not revenue. Estimate active players, sessions and average stake, and let wagering volume fall out.
2. Apply a defensible hold rate per vertical. Anchor on a published, derivable figure and record its source and period.
3. Deduct gaming duty at the statutory rate, on the correct base, checking whether bonuses are dutiable.
4. Deduct bonus cost at wagered-through value, not granted value, modelling the two separately.
5. Deduct payment cost as a percentage plus a fixed fee, applied to your deposit size distribution, not an average.
6. Apply platform and content revenue share to the resulting NGR, using the contract's own definition and any monthly minimum.
7. Subtract acquisition at a CPA your channel mix supports, then compute payback per cohort. If payback exceeds player lifetime, the model fails at any volume.
8. Stress the result. Re-run with duty raised by half, hold down by your vertical's volatility band, and payback a third longer. A model that only works at its central case is not a model.

Step eight is not academic. Operators modelling on a 21% UK duty rate in early 2026 were running 40% by April. Sportsbook hold moved more than five points across eighteen months, and any sportsbook attached to a casino inherits that variance.