How we score platform providers
Every provider is assessed on ten published axes and rolled up with explicit category weights. This is methodology v2.0: the formula is code-backed, versioned, and independent of paid placement.
Last updated August 20, 2026
We version this methodology so each review can cite the exact rubric used. A methodology change requires a version bump and a recalculation of every overall score.
Our independence
- Providers cannot pay for a higher score or a better rank.
- Lead delivery is currently deferred; any future referral fee must never move a score.
- Every provider is scored by the same rubric, not by who works with us.
- Paid placements (featured or sponsored) are labeled and kept visually separate from editorial scores.
- We review providers whether or not they have a commercial relationship with us.
More on the business model is on the how we make money page, and our standards are in the editorial policy.
The ten axes and their weights
The overall score is a weighted roll-up of the ten axes below. Sportsbook is scored as N/A (not zero) when the current assessment does not record a sportsbook product. An unknown state is therefore not converted into a zero or an unsupported conclusion.
The weights below are the source of truth. They are editorial policy, not values fitted to preserve previously published overall scores. Every category sums to 100%.
| Axis | Packaged platform | SaaS / PAM | Game aggregator | Sportsbook platform | Crypto-casino platform | Sweepstakes / social |
|---|---|---|---|---|---|---|
| Game content & aggregation Catalog scope, supplier breadth, owned content, and product availability. | 13% | 10% | 30% | 3% | 15% | 20% |
| Payments & crypto Payment integrations, crypto support, currencies, and evidenced payment operations. | 12% | 13% | 5% | 8% | 24% | 7% |
| Sportsbook Engine ownership, trading model, coverage, and operational control. | 8% | 8% | 0% | 30% | 8% | 3% |
| Bonus & CRM Bonusing, segmentation, retention, and engagement tooling. | 10% | 12% | 8% | 8% | 8% | 15% |
| Compliance & licensing Credential scope, KYC, geolocation, responsible-gambling controls, and evidence quality. | 14% | 14% | 16% | 15% | 14% | 25% |
| Time to market A realistic launch range and the work that drives it. | 14% | 8% | 10% | 8% | 10% | 8% |
| Pricing & value Transparency and value for the evidenced product scope. | 10% | 7% | 10% | 8% | 8% | 7% |
| Customization APIs, modularity, front-end flexibility, and integration control. | 6% | 12% | 8% | 7% | 5% | 5% |
| Tech & reliability Architecture, scale, measured uptime, incidents, and recovery evidence. | 7% | 10% | 8% | 8% | 5% | 6% |
| Support Support coverage, account management, SLAs, and managed services. | 6% | 6% | 5% | 5% | 3% | 4% |
If an axis is genuinely not applicable, its weight is removed from the denominator rather than converted to zero. For example, a content-only aggregator's sportsbook axis carries zero weight for that category, so it does not affect the result whether or not a sportsbook product is recorded.
How to compare category-weighted scores
Compare overall scores within the same provider category first. The underlying axis scale is shared, but the overall weights are not: an aggregator score emphasizes content and integration, while a sportsbook score gives more weight to betting, trading, and reliability. An 8.0 in one category therefore does not prove broader fit than a 7.8 in another.
For a cross-category shortlist, compare the relevant axis scores and evidence directly, then apply the buyer's non-negotiable requirements. A category average describes the recorded peers in that category; it is not a market benchmark or a substitute for product, jurisdiction, and contract diligence.
Public overall scores are rounded to one decimal for display. Ranked lists use the same weighted calculation before that rounding, so two providers shown as 8.2 do not fall back to an arbitrary alphabetical or file order. The stable slug order is used only when the unrounded calculated results are equal.
- Use the procurement-scope selection matrix to define the award boundary and non-negotiable acceptance gates.
- Use the platform finder to define delivery model, vertical, market, and control requirements.
- Use the side-by-side comparison to inspect the same axes across the resulting shortlist.
- Move eligible candidates into the buyer-owned RFP for scoped evidence, testing, economics, and contract schedules.
Publication standard
Every published fact is bounded by the exact legal entity, product, market, period, status, and definition it supports. Profiles state the resulting position directly; the supporting record remains in the internal data layer for audit and refresh.
Certainty is explicit: verified, unverified, estimated, unknown, or contract-only. Missing support never becomes a settled licensing, capability, deployment, or contract conclusion.
Volatile catalogue, integration, payment, language, event, launch-time, uptime, client, award, and deployment figures enter comparable fields only with a dated scope and stable definition. A named deployment does not establish uptime, support quality, commercial success, or a broader product relationship.
Quantitative product copy is accepted only as an atomic, scoped claim. A changed number, a repeated number in a different context, or a claim without its matching audit record does not enter the public picture.
The product and module register publishes the resulting controlled product relationships. A product name does not become package inclusion, buyer entitlement, certification, or market access.
The provider capability matrix applies the same ten provider-level questions to every profile. A Recorded cell identifies a bounded capability state, Not confirmed remains an open question, and silence never becomes an unsupported or unavailable conclusion.
Every structured fact retains its supporting record in the data layer. Validation fails when identity, licensing, product, pricing, reliability, or scoring data loses that audit trail.
An estimated implementation range is our planning band, not observed delivery performance or a supplier promise. The band must name the product context and material dependencies; buyers should replace it with a dated implementation plan, milestones, acceptance criteria, and remedies for their proposed project.
The implementation plan replaces generic launch timing with dependency owners, work products, acceptance conditions, hold conditions, and go/no-go decisions.
How we handle licences and approvals
We do not convert a brand-level phrase such as "licensed in" into blanket market coverage. A credential record must identify the legal entity, market, credential kind, status, product scope, evidence state, and checked date. Only an effective, active record backed by qualifying evidence contributes to recorded coverage. Pending, unknown, inactive, expired, assistance-only, and technical-certification records do not become market coverage.
A supplier or product approval does not replace the operator licence, market access, game or wagering approval, payment approval, or other launch requirements. Published validity and renewal checks are recorded separately, and a pending renewal or an unsuccessful search does not silently create a new credential term.
Jurisdiction dossiers keep permission roles separate. The Malta B2B supplier register distinguishes critical-supply permissions from B2C operating contexts, product acceptance, and production deployment.
The regulatory coverage matrix compares recorded-active relationships across markets without turning profile relationships or conservative record groups into licence totals. The Ontario exact register additionally separates B2B supplier registrations from site-level operator registrations.
Conflicts and scope
Apparent conflicts are separated by entity, product, jurisdiction, definition, period, and date before comparison. Legal status follows the exact authority record; product and operational facts keep their own narrower scopes.
If a conflict remains, the comparable value is unavailable. Provider pages publish that final boundary instead of narrating competing claims or selecting the more attractive figure. A newer record is not stronger when it covers a different entity, product, or measurement method.
How we treat reliability evidence
Reliability percentages are comparable only when they describe the same service scope, observation period, calculation method, exclusions, and incident rules. We therefore separate independently observed results, first-party status telemetry, complete service-level terms, and marketing claims instead of presenting every uptime percentage as the same kind of fact.
An incident count means only the recorded objects in the stated window. It is not a provider-to-provider failure rate. An empty recorded history is not proof that no outage occurred, and a status page without incident narratives may be incomplete even when component states show disruption. We calculate a recovery average only when every included record has exact start and resolution timestamps and the records cover one comparable service scope.
Architecture descriptions, certifications, client logos, company size, and transaction volume can shape due-diligence questions, but none of them is measured uptime. A genuine public service-level target must disclose the measurement window, calculation method, exclusions, and breach remedy; otherwise an SLA-labelled percentage remains a unverified marketing claim in the data layer.
How scores are set
Axis scores are our editorial assessment, informed by product depth, evidence quality, and track record. The public overall is a deterministic weighted calculation from those axes. Architecture may suggest questions about ownership or portability, but it is not evidence of contract terms; unsupported ownership, exit, minimum-term, migration, and exclusivity values are shown as contract terms not confirmed.
What we don't do
- We don't take payment to change a score or a ranking.
- We don't publish star-rating / aggregate-rating markup designed to game search results.
- We don't list game studios as platform providers. Studios appear only where we explain the difference.
Freshness & corrections
Each review shows when it was last reviewed. We re-check fast-moving fields (licensing, pricing, and public reliability records) by priority. The contact status page explains why correction submissions are currently paused.