BonusBandit Trust Score · Methodology v1.0
A 0-100 transparent, evidence-based score. Published methodology, source-tagged data, no affiliate bias.
Every other number on this site is in service of that one question. A sweepstakes casino can have a slick app, a big welcome bundle, and friendly chat agents and still be a place you should not deposit your time into — because none of that decides whether a legitimate redemption actually lands in your account. The Trust Score is built to answer the question a player is really asking, and to answer it the same way for every operator so the scores are comparable.
We score the operator, not the offer. A good bonus on an operator that stalls payouts is not a good deal. The score puts payout reality first and treats everything else as context.
Factors that determine whether a player gets paid — redemption and KYC — dominate the score. Pleasant support that never pays out is worse than gruff support that always pays. About half the total weight sits on the two "do you actually get paid" factors. That is intentional and correct for this niche.
A weighted average alone is dishonest, because a fatal flaw can be masked by nice-to-haves. Certain catastrophic, documented findings cap the maximum score regardless of everything else. A casino that systematically refuses legitimate redemptions cannot buy its way back with fast support and a clean app.
Every factor is rubric-based, so it is countable and auditable. Every data point is source-tagged. The methodology is published — this page. When we are missing data, we lower a visible confidence rating instead of quietly guessing a number that looks more precise than it is.
| # | Factor | Weight | Why it carries this weight |
|---|---|---|---|
| 1 | Redemption reliability | 30% | The core trust question: do they pay, how fast, and do they pull tricks to avoid it? Nothing else matters if the money never arrives. |
| 2 | KYC fairness | 20% | Identity checks are the most commonly weaponized stalling tactic used to avoid paying. How fairly KYC is applied is a payout question, not a paperwork question. |
| 3 | ToS honesty | 15% | Scores the published terms against a checklist of red-flag clauses. Objective and legally safe because it reads what the operator already wrote down. |
| 4 | Playthrough fairness | 12% | Wagering requirements on redeemable coins decide how much of a "win" you can actually cash. Hidden or shifting playthrough quietly erases value. |
| 5 | Operator track record / ownership | 12% | Who runs it, how long they have operated, and how their other brands behave. The differentiator most review sites skip. |
| 6 | Reviews (adjusted) | 7% | Useful sentiment, but heavily gamed — so weighted low and corrected for review farming. Resolution of complaints beats raw star counts. |
| 7 | Support responsiveness | 4% | A leading indicator, deliberately low-weight. Standardized with the same test query across operators so it is comparable. |
This is the heart of the score. We look at the share of documented redemption requests that complete without unjustified denial, the median time it takes to get paid, how high the minimum redemption threshold is (lower is fairer), and any documented pattern of canceling or reversing valid redemptions. A casino that pays nearly everyone within two days, lets you cash out at a low threshold, and has no reversal pattern scores high. One that denies legitimate cashouts, drags payouts past three weeks, or quietly reverses approved redemptions scores low.
Identity verification is normal and required. Weaponized KYC is not. We reward operators that make verification clear and doable up front, ask for a reasonable document set (ID and proof of address), and process checks quickly. We penalize operators that spring KYC only at the moment of withdrawal as a stall, demand excessive or invasive documents, or have documented cases of using KYC to deny or delay legitimate payouts.
This is a pure checklist over the operator's own published terms. A page starts at full marks and loses points for each red-flag clause present: broad balance-forfeiture rights, a low maximum-cashout cap, undisclosed or unclear wagering requirements, retroactive term-change rights without notice, a missing or unreasonably burdensome no-purchase (AMOE) path, dormancy or inactivity fees, and predatory prize-forfeiture conditions. Every deduction links to the exact clause and the date it was captured.
Wagering requirements on redeemable Sweeps Coins decide how much of a notional win you can actually withdraw. A clearly stated 1x requirement scores well; a high multiplier, or one that is never clearly stated, scores at the floor. We also check whether playthrough is disclosed before you play rather than buried, whether free coins carry undisclosed wagering, and whether requirements have changed mid-stream.
We look at the reputation of the parent or operator family (derived from how the owner's other brands score and behave), how long the operator has run, any documented disputes or exit-scam history, and whether there is an identifiable legal entity with a disclosed jurisdiction. Opaque ownership and a thin history pull this factor down; a clean, transparent, long-running operator pulls it up.
Star ratings are easy to game, so we weight them low and correct them. The biggest component is the complaint-resolution rate — what share of formal complaints actually got resolved acceptably for the player. We volume-adjust Trustpilot scores and flag farming patterns, and we read structured community sentiment from recent threads. We always store the underlying counts and the capture date, not just a final number.
We send the same standardized test query to every operator and measure the median first-response time, the channels available (live chat, email, phone), and whether the query was actually resolved. It is a useful leading indicator, but it is deliberately the lightest factor: fast, friendly support that does not translate into payouts is not what the score is protecting.
A weighted average alone is dishonest. If one fatal flaw can be averaged away by a stack of minor strengths, the number lies about how risky the operator is. So after the weighted sum, we apply non-compensatory gates — kill switches that cap the maximum possible score when a catastrophic, documented condition is present. The final score is the lower of the weighted score and the strictest applicable cap.
Documented, systematic refusal of legitimate redemptions caps the score at 25. This is the worst thing a sweepstakes casino can do, so no other strength can lift it out of "Avoid".
KYC documented as systematically used to deny legitimate payouts caps the score at 30.
An exit scam or mass non-payment event on a sibling brand from the same operator family caps the score at 35.
Operating with no identifiable legal entity and undisclosed ownership caps the score at 40.
Every triggered gate attaches its linked, dated evidence and surfaces a plain-language reason on the operator's page — for example, "Score capped: documented redemption denials — see sources." The cap is the honesty mechanism, not a punishment.
The Trust Score sits next to a separate confidence rating shown as High, Medium, or Low, plus a "Last verified" date. Confidence never changes the Trust Score — it contextualizes it. It is built from three things: how much of the total weight is backed by real data (coverage), how recent that data is (recency, which decays as a factor ages past its expected refresh interval), and how strong the sources are (tested observations count fully, aggregated-only data counts at a discount).
This is the opposite of fake precision. If we have run a redemption ourselves and re-checked the terms last week, confidence is High and the number is trustworthy. If half the factors are estimated from old third-party complaints, confidence is Low and the score says so out loud. Admitting what we do not yet know is itself a trust signal.
The score is the asset the whole brand hangs off, and the fastest way to destroy it is to let affiliate revenue bias the ratings. So the order is fixed: we score first and monetize second, and an affiliate link can appear only on an operator that already rates well on its own merits. A commercial relationship can change which link a page uses; it can never change the risk language, the factor scores, the gates, or the confidence rating. Read the affiliate disclosure for the compensation policy that pairs with this methodology.
Every data point is tagged with where it came from, so a reader can judge it and an editor can re-check it.
This methodology is version 1.0. We expect to refine the weights, rubric thresholds, and gates as we learn what actually predicts a bad payout experience. When that happens we bump the version and recompute history, so a score that changes is always explainable: you can see whether the operator changed, the data changed, or the methodology changed. Weights, thresholds, and caps live in a versioned configuration file separate from the scoring code, and raw inputs are stored apart from computed scores so any past score can be reproduced.
Because the one question a player actually cares about is whether they can get their money out. Redemption and KYC together carry about half the score because they decide whether a payout happens at all. Pleasant support or a generous-looking bonus cannot make up for a casino that does not pay.
Nobody pays to be scored, and no operator can pay to change a score. Affiliate placement is decided after a score is set and never feeds into it. We score first and monetize second, and only on operators that rate well on their own merits.
Yes. Every score traces to dated, linked evidence, and operators have a visible right of reply. If an operator shows that a fact is wrong or out of date, we correct the underlying data, recompute the score, and log the change with its date.
Cadence depends on the factor. Aggregated reviews refresh roughly weekly, published terms and ownership are re-checked about quarterly or whenever a change is detected, and tested factors like redemption and KYC run on a scheduled test cycle or sooner if complaints spike. Older data lowers the visible confidence rating.
A cap is a non-compensatory gate. Certain catastrophic, documented findings — such as systematic refusal of legitimate redemptions — limit the maximum score no matter how strong the other factors are. A weighted average alone is dishonest because a fatal flaw can be hidden by nice-to-haves, so we cap instead.
Confidence is a separate High, Medium, or Low rating shown next to the score. It reflects how much of the methodology is backed by real, recent, high-quality data. Missing or stale data lowers confidence rather than being guessed, so the number never pretends to be more precise than the evidence behind it.
New operators are not punished into a fake-low score. Their longevity sub-metric scores low and they are labelled Unproven, with reduced confidence. A new brand from a strong operator family can still inherit a reasonable family-reputation sub-score.
Probably, as we learn. The methodology is versioned, starting at version 1.0. When weights, rubrics, or gates change, we bump the version and recompute history so every score change is explainable rather than silent.
We are publishing the methodology before scoring a single operator on purpose. The published method is the promise; operators get scored only when the data behind each factor is rigorous enough to stand behind. First operator scores publish on a rolling basis. Spot a factual error or want to submit a correction? Reach the editorial team at noah.rafkin@bonusbandit.win or through the about page.