Our methodology in one page
The money page is the tightest summary of Meridian's scoring approach that we publish. Aisha Rowland wrote the first version of this rubric during a two-week desk retreat and it has been refined roughly quarterly since. This section is the compact version. The homepage carries the full narrative, and the individual columns are unpacked below.
The four principles behind the rubric
The rubric is built on four principles. First, every dimension must be measurable in a way that survives peer review – we should be able to hand the score to another journalist and have them reproduce it within one score-point. Second, every dimension must be independent enough of the others to add real information rather than covariate signal. Third, the weights must be defensible on grounds that are not the outputs the weights produce – we do not tune the weights to make particular brands rank particular ways. Fourth, the composite must be a public number, expressed on a hundred-point scale, and dimensional scores must be published alongside it so readers can see where the composite comes from.
How we score each dimension
Every dimension returns a zero-to-five score after normalisation. Zero is the worst observed value in the sample, five is the best. For continuous variables such as withdrawal SLA, we bucket into five bands using an even quintile split across the current sample. For categorical variables such as licence tier, we assign the ordinal score directly. For hybrid variables such as terms readability, we compose two subscores (grade level and clause specificity) into a single dimension using an equal-weight combination. The dimensional scores are then multiplied by their weights and summed to produce the composite. The whole calculation is transparent and reproducible from the published dimensional scores.
The role of measurement drift
Between sample refreshes, brands move. New welcome bonuses launch, withdrawal SLAs speed up or slow down as the operator's KYC workflow shifts, and licence tiers occasionally change under regulatory reforms such as the Curacao LOK reform. We handle this with a changelog on the money page, and every published scorecard is stamped with its measurement date. Where a brand has moved sharply since the last refresh, we flag it in the changelog and, if the move is large enough, we run an interim rescore rather than waiting for the next full refresh. Aisha Rowland has been explicit on the desk that transparent handling of drift is more important than pretending our sample is real-time.
The composite index explained
The composite index is the single number we publish for each brand in the sample. It is a weighted sum of the twenty-two dimensional scores. This section explains how the composite is calculated, how it should be read, and what its limitations are.
The mathematics of the composite
Composite is calculated as sum of (dimensional score × dimensional weight × twenty), where the dimensional score is on a zero-to-five scale and the dimensional weight is a percentage summing to one hundred across the twenty-two dimensions. Multiplying the score by twenty converts the zero-to-five band into a zero-to-one-hundred contribution scaled by the weight. The composite therefore sits on a zero-to-one-hundred scale, with a theoretical maximum of one hundred (every dimension at its top-of-sample level) and a minimum of zero (every dimension at its bottom-of-sample level). No brand in our sample has ever posted a composite above eighty-two or below forty-nine.
Reading the sample distribution
The sample distribution is right-skewed with a long left tail. The median composite across the twenty-two brands is around sixty-two. The top quartile begins at seventy. The bottom quartile ends at fifty-six. The interquartile range is fourteen points, which reflects the substantial genuine variation across the sample. Where a brand sits above seventy composite, that puts it in the top quartile, and a brand above seventy-five is in the top decile. Where a brand sits below fifty-five composite, that is bottom-quartile territory. Between fifty-five and seventy, the sample is dense enough that composite gaps of less than three points should be read as ties rather than as ordinal separations.
What the composite does not tell you
The composite is a summary statistic and, like any summary statistic, it hides more than it reveals. Two brands at composite seventy can have very different dimensional profiles. One might be a tier-two Kahnawake brand with heroic scores on withdrawal SLA and support but a weaker welcome bonus. Another might be a tier-four legacy Curacao brand with a strong welcome bonus but weaker structural scores. Both sit at seventy composite. Neither is a straightforward substitute for the other. The composite is a useful sorting mechanism, and it is a useful topline for a busy reader, but the dimensional profile is where the meaningful editorial content lives. We publish both. Read both.
| Composite band | Sample count | Interpretation |
|---|---|---|
| 75+ | 2 | Top decile of sample |
| 70 to 74 | 4 | Top quartile |
| 62 to 69 | 7 | Sample median band |
| 56 to 61 | 5 | Below median |
| Below 56 | 4 | Bottom quartile |
The licence-tier column
Licence tier is the heaviest single-column weight in the composite and the least noisy variable in the sample. This section explains how the tier column is constructed and how it plays into the overall ranking.
The four tiers again, in operational terms
Tier one, awarded to MGA, Gibraltar and Isle of Man licences, carries the highest score because those jurisdictions have the deepest published supervision regimes and the most tested complaint routes. Tier two, Kahnawake, carries the second-highest score because the KGC's dispute framework has visible enforcement history. Tier three, post-LOK Curacao B2C licences issued under the new National Ordinance regime, is the mid-band because the reform has yet to accrue a substantial enforcement track record but the framework is credibly more supervised than the previous master-and-sub structure. Tier four, legacy Curacao and Anjouan, occupies the bottom band because the effective supervision, as measured by publicly observable complaint outcomes, is thin.
Where the tier column matters most
Because licence tier is the heaviest single column and the least noisy, it does more work than any other individual variable at the composite margins. A brand can gain up to four composite points by moving from tier four to tier one, and no other single dimension has that swing. That is deliberate. Aisha Rowland's argument, articulated in the desk memo, is that licence tier is the closest thing we have to a proxy for consumer-protection quality, and its heavy weighting is what makes the composite behave like a consumer-protection index at all. If we weighted licence tier at say four percent instead of ten, the composite would be dominated by welcome-bonus size and would reward marketing over structure.
Verifying a brand's licence status
Every brand in our sample is checked against the register of its licensor. For Curacao, that means the Curacao Gaming Control Board register at gcb.cw. For Anjouan, the Anjouan Gaming Authority register. For Kahnawake, the Mohawk Council of Kahnawake register. For MGA, the authorised operator search. For Gibraltar, the Gambling Commissioner's list. Where a brand's public licence claim does not resolve to a live register entry, we exclude the brand from the sample. This has happened twice in the last four refresh cycles, both times because the brand had allowed its licence to lapse without updating the site.
The welcome-bonus column
The welcome bonus is where operators compete hardest in marketing terms, and it is where the noise-to-signal ratio in the sector is at its worst. Meridian's approach separates the headline from the effective value. This section walks through the column construction.
Headline value, band-by-band
The headline value is the maximum bonus GBP-equivalent, capped at whatever the operator publishes. Across our sample the bands are as follows. Under two hundred and fifty pounds is the small band, populated by four brands. Two hundred and fifty to five hundred pounds is the low-mid band, populated by six brands. Five hundred to one thousand pounds is the sample-modal band, with eight brands. One thousand to one thousand five hundred is the large band, with three brands. Above one thousand five hundred is the top band, with one brand publishing a headline of just under two thousand pounds. The headline correlates only weakly with the composite, because headline-heavy brands tend to accompany the large headline with heavier wagering multipliers.
Effective value after wagering probability
Effective value is what we calculate as the expected withdrawable value from the offer, assuming the bonus is fully consumed at the sample-modal RTP of ninety-six percent. The formula deducts the expected turnover-driven loss from the headline. On the sample-modal offer of a five-hundred-pound match at thirty-five-times deposit-plus-bonus, effective value comes out at around two hundred and twenty pounds. On the top-band offer of a two-thousand-pound match at forty-times, effective value comes out at around six hundred and fifty. So the top-band operator does still offer more expected value than the modal one, but not four times more. The headline overstates the differential by a factor of two.
No-wager offers and their real cost
Four brands in the sample publish a no-wager welcome stream. The headline on these offers is markedly smaller – typically fifty to a hundred and fifty pounds – but the effective value is close to the headline, because the wagering-driven loss is zero. On expected value terms, a hundred-pound no-wager offer is materially better than a five-hundred-pound offer at thirty-five-times. This is not a widely appreciated point in the sector's marketing, and it is one of the pieces of arithmetic the money page exists to make legible. Aisha Rowland has argued repeatedly that no-wager offers are the honest end of the promotional space, and their scoring in our rubric reflects that view.
| Offer type | Headline (median) | Wagering | Effective value (est.) |
|---|---|---|---|
| Small match | GBP 200 | 30x | GBP 105 |
| Modal match | GBP 500 | 35x | GBP 220 |
| Large match | GBP 1,200 | 40x | GBP 460 |
| No-wager stream | GBP 100 | 1x | GBP 96 |
The withdrawal SLA column
Withdrawal SLA is the single most consistent predictor of user sentiment across our sample. It is also one of the more variable columns, because the SLA depends on the rail, on the operator's internal workflow, and on whether the KYC gate has already been cleared. This section unpacks the column.
First-withdrawal SLA distribution
Measured first-withdrawal SLA across our sample runs from around six hours at the fastest brand (a Kahnawake-licensed crypto-first operator with same-day KYC review) to over five business days at the slowest (a legacy-Curacao brand with a manual bank-transfer workflow). The sample-median first-withdrawal SLA is roughly forty-eight hours, and the interquartile range is twenty hours to seventy-two hours. First-withdrawal SLA correlates strongly with licence tier – tier-one and tier-two brands cluster in the sub-twenty-four-hour band, while tier-four brands are much more evenly spread across the whole range.
Steady-state SLA distribution
Steady-state SLA, measured on a second or subsequent withdrawal once KYC is cleared, is meaningfully faster. The sample-median steady-state SLA is around eight hours, with an interquartile range of three hours to eighteen hours. The rail explains most of the remaining variance – crypto rails settle in minutes plus network confirmation, e-wallets in a few hours, cards in a day, bank transfer in a few days. This is where the operator's own processing window matters – the difference between a two-hour operator processing window and a twelve-hour operator processing window can be the difference between a same-day withdrawal and a next-day withdrawal on an e-wallet rail.
Weekend and holiday effects
Where the operator processes withdrawals only during working hours, weekend and holiday effects can extend SLAs substantially. Nine of our twenty-two sample brands have visible weekend gaps in their processing workflow. A Friday-evening withdrawal at one of those brands can take until Monday afternoon. This is not always disclosed in the advertised SLA. We measure and report it separately, and we score down brands that carry a large weekend gap without acknowledging it in the terms. Aisha Rowland has written that the weekend gap is one of the sneakier variables in the sector, because it disproportionately affects players who happen to hit a good result during a weekend session, and those are the players who most notice the delay.
| Rail family | First-withdrawal (median) | Steady-state (median) | Sample support |
|---|---|---|---|
| Crypto | 18 hrs | Under 1 hr | 21 / 22 |
| E-wallet | 36 hrs | 4 hrs | 18 / 22 |
| Card (Visa Direct) | 48 hrs | 24 hrs | 14 / 22 |
| Bank transfer | 72 hrs | 2 days | 10 / 22 |
The game library column
Game library scoring combines slot count, live dealer provider mix, and supplier diversity into a single normalised column. This section walks through how the column is constructed and what it does and does not reflect.
Slot count on a log scale
Slot count is scored on a log-base-two scale to prevent the top-end brands from dominating the column. Under eight hundred titles scores one. Eight hundred to sixteen hundred scores two. Sixteen hundred to thirty-two hundred scores three. Thirty-two hundred to sixty-four hundred scores four. Above sixty-four hundred scores five. Across our sample the modal band is three, corresponding to twenty-two brands whose slot counts sit in the sixteen-hundred-to-thirty-two-hundred band. Log scaling reduces the marketing incentive to inflate slot counts through supplier-catalogue padding, because the marginal score gain from adding another supplier's back catalogue is small once you are already in a mid band.
Live dealer provider score
Live dealer is scored on a five-band scale by studio coverage. One point for the presence of any live dealer offering. Two points for at least one major studio (Evolution, Pragmatic Play Live, Playtech Live). Three points for Evolution specifically. Four points for Evolution plus at least one other major studio. Five points for Evolution plus at least two other major studios. This weighting reflects Evolution's dominance in the live segment and the practical reality that a live catalogue without Evolution is a meaningfully thinner catalogue. Seventeen of our twenty-two brands carry Evolution and therefore score three or above on this sub-column.
Supplier diversity index
Beyond slot count and live provider score, we compute a supplier diversity index using the Herfindahl-Hirschman formulation applied to the operator's supplier catalogue. Brands with concentrated catalogues – where a single supplier accounts for more than forty percent of titles – score lower on this sub-column because catalogue concentration is a supply-side fragility. If the dominant supplier's terms shift or their catalogue is removed from the operator, the catalogue thins abruptly. Diversified operators, where no supplier exceeds twenty percent of titles, score higher. The three components combine into the final game library column at forty percent slot count, thirty percent live dealer, thirty percent supplier diversity.
The payment coverage column
Payment coverage rolls the deposit-method count, withdrawal-method count, and payment rail balance into a single column, with an adjustment for whether the operator supports low-friction crypto rails. This section explains the calculation and what it captures.
Deposit and withdrawal method count
Method count is a simple integer, but it is not the whole picture. A brand with fifteen deposit methods including six near-duplicate wallet options scores worse than a brand with eight deposit methods covering card, e-wallet, crypto, bank transfer and Paysafecard cleanly. We adjust the raw count by a category-coverage bonus. Full category coverage – at least one method in each of card, e-wallet, crypto, bank transfer – adds one point to the raw count. Partial coverage (three of four categories) adds half a point. Coverage in only two categories or fewer receives no bonus and can attract a deduction. Across our sample, fifteen of twenty-two brands cover all four categories on deposit.
Crypto-support depth
Crypto support is scored not by token count but by chain count. Supporting Bitcoin, Ethereum and one Ethereum stablecoin (USDT-ERC20) scores as three tokens but only two chains, because the stablecoin lives on Ethereum. Supporting BTC, ETH, USDT-ERC20 and USDT-TRC20 scores as three chains, because Tron is a distinct chain. Chain diversity matters for withdrawal cost, because Ethereum network fees are meaningfully higher than Tron or Bitcoin Lightning. Sample-median chain count is three. The best-in-sample brand supports six chains including Solana and Polygon alongside the core three. Chain count contributes to the payment coverage column at a low weight because most players do not use crypto, but where an operator does support diverse chains, that signals a payments team that has done its homework.
Withdrawal-method asymmetry
An often-overlooked payments variable is the asymmetry between deposit and withdrawal methods. A brand can accept card deposits but only support e-wallet or bank transfer for withdrawal, because Visa Direct and Mastercard Send are less broadly available and more expensive than card acquiring. Where an operator's deposit-to-withdrawal method mapping requires the customer to withdraw via a different rail than they deposited, this creates friction, particularly on first withdrawal when the destination rail may need separate verification. We flag operators with more than a two-method-count gap between deposit and withdrawal, and this contributes a small deduction to the payment column.
A worked scoring example
The best way to understand the scoring is to see it applied to a concrete brand profile. This section works through a stylised example, using a brand profile constructed from the sample-modal values on each dimension. The brand is not a real operator, but each value is drawn from the sample distribution.
The stylised sample-modal brand
Our stylised brand holds a legacy-Curacao licence (tier four), publishes a five-hundred-pound match welcome bonus at thirty-five-times deposit-plus-bonus wagering, carries around four thousand slots including Evolution live dealer, has a first-withdrawal SLA of forty-eight hours and a steady-state SLA of eight hours, sets its KYC threshold at two thousand five hundred pounds cumulative deposit, accepts card, e-wallet, crypto (three chains) and Paysafecard on deposit and card, e-wallet and crypto on withdrawal, has 24/7 advertised support with a twenty-minute median response, scores five out of nine on the mobile UX rubric, offers a core RG toolkit but not extended tools, and has terms and conditions at Flesch-Kincaid grade fifteen with four of six critical clauses stated numerically.
Column by column scores
Licence tier scores one (tier four). Welcome bonus size scores three (modal band). Wagering multiplier scores three (modal thirty-five-times). Game count scores three (sixteen-hundred-to-thirty-two-hundred band on log scale). Live dealer scores three (Evolution present). Withdrawal SLA scores three (sample-median first-withdrawal). KYC threshold scores three (sample-median value). Payment coverage scores four (full deposit coverage, three-chain crypto, minor withdrawal asymmetry). Mobile UX scores three (five of nine sub-checks passed). Support scores three (24/7 advertised and mostly staffed). RG toolkit scores three (core present, extended absent). Terms readability scores three (grade fifteen, four of six clauses numeric). The remaining ten dimensions all score close to sample median at around three.
The composite calculation
Multiplying each dimensional score by its weight and by twenty, then summing, produces a composite of just under sixty-two for our stylised modal brand. That is right at the sample median. This gives a sense of how the rubric behaves on a typical mid-sample profile. To move the composite substantially upward, the brand would need to move on one or more heavy-weight dimensions – licence tier upward, withdrawal SLA downward, or terms readability upward. A one-tier improvement on licence would add roughly two composite points. A twenty-four-hour improvement in first-withdrawal SLA would add roughly one and a half. A grade-two improvement in terms readability would add just under one. The mechanics of moving the composite are transparent, and the rubric rewards structural improvements over marketing improvements.
- 1Assign the licence tier. Verify licence claim on the licensor register and assign tier one to four.
- 2Extract structural values from the terms. Bonus size, wagering multiplier, KYC threshold, deposit and withdrawal minimums.
- 3Test the payment coverage. Attempt a small test deposit on card, e-wallet and crypto; note the deposit and withdrawal method counts.
- 4Measure the withdrawal SLA. Run a first-withdrawal test and record measured versus advertised.
- 5Score the experiential dimensions. Mobile UX rubric, support hours and first-response, terms readability index.
- 6Compute the composite. Multiply each score by its weight and by twenty, sum to a zero-to-one-hundred composite.
What our scoring does not tell you
A scoring rubric is only as useful as the honesty of its own limits. This section is where we lay out the things our rubric does not measure, does not claim, and cannot substitute for.
The rubric does not measure individual fit
The composite is a summary across the sample, not a recommendation for any individual reader. Whether a particular brand fits your play patterns depends on things the rubric does not capture – the specific games you play, your typical stake sizes, your usual session lengths, your preference for slot volatility. If you are a low-volatility slots player at ten-pence stakes, some of the twenty-two dimensions matter more to you than they do to a high-volatility crash-game player at ten-pound stakes. The rubric is a starting point for a decision. It is not the decision itself.
The rubric does not measure future performance
All scoring is retrospective. It reflects the operator as it was on the day of measurement. Between refreshes brands can and do change. Licence conditions shift. Ownership changes. Payment providers withdraw or reintroduce services. Support quality drifts as team turnover shifts. Our scoring is a snapshot with a date stamp, not a forecast. Where a reader relies on the score for a decision, they should also check the changelog for material changes since the sample refresh, and they should treat the score itself as one input among several.
The rubric does not measure solvency risk
The most consequential thing the rubric does not measure is operator solvency. If an offshore operator becomes insolvent, customer balances may or may not be recoverable, depending on the licence jurisdiction's segregated-funds requirements. Curacao's post-LOK regime requires segregation, but enforcement is thin. Anjouan and legacy Curacao do not require segregation. MGA, Gibraltar and IoM require it and enforce it. This is another reason licence tier gets heavy weight in the composite – it correlates with solvency-risk reduction – but the composite itself does not directly measure solvency risk, and no reader should treat it as an insolvency prediction. Where you keep meaningful balances at an offshore operator, keep them small.
Responsible gambling and money-page ethics
A money page is where the pull toward recommending grows strongest, and where the editorial discipline needs to be tightest. Meridian's money page ethics are set out in this section, alongside the standard responsible gambling signposting that runs across every content page on the site.
Why we do not name recommended operators
Naming a specific operator as a recommendation would require us to underwrite that operator's ongoing quality, and we do not have the resources or the mandate to do so. What we do is publish the dimensional scores and let the reader compose their own ranking. If a reader concludes from our data that brand X is the best fit for them, that is a reader's conclusion, not a Meridian recommendation. This is not a semantic distinction. It is an editorial one, and it changes the incentive structure of the whole publication. Aisha Rowland has said publicly that this is the single decision she is proudest of on the desk.
Support signposting on every page
Every content page on Meridian carries the GamCare 0808 8020 133 signpost, alongside references to BeGambleAware, Gamban and BetBlocker. This is not decoration. It is a working editorial policy that responds to the practical reality that some readers arriving at pages like this one are already at a point where the signposting is useful. If you are one of those readers, the helpline number is above, it is free, it is confidential, and it is staffed twenty-four hours. There is no requirement to identify yourself. There is no minimum threshold of concern. The helpline exists for exactly the kind of ambiguous half-worry that most people have before they call it a problem.
What we would say to a friend
If a friend came to us asking whether they should open an account at an offshore operator, our honest answer would be, it depends on why. If the reason is a considered preference for higher deposit ceilings, faster crypto rails, or a bigger welcome offer, and if the play sits comfortably inside a household budget the person can sustain losing, then the choice is theirs and our data can help. If the reason is a growing sense that the UKGC's frictions – limits, self-exclusion, KYC – are inconveniences to be routed around, that is the sense we would ask the friend to sit with for a while before opening the account. The frictions exist because self-exclusion is a proven harm-reduction tool. Routing around them is a decision most people go on to regret.