Canonical: https://protectproduct.com/methodology/ [Home](/) / Methodology Protect Product / Field guide # How we score fraud detection tools The editorial rubric contains 22 weighted criteria totaling 100 percent. Here is how to use it, and where its limits are. Read each score alongside the product’s use case, sources and practical trade-offs. ## Scope and eligibility The comparison includes products that identify visitors, devices or IP addresses and help detect risky users. It excludes products focused only on CAPTCHA, KYC, chargeback guarantees or a general web application firewall. Some included tools specialize in only one layer of protection; the category is intentionally broad so buyers can see specialist and platform trade-offs. ## Editorial scoring Each criterion has a weight; the 22 weights add to 100 percent. Scores out of ten express an editorial judgment of overall buyer fit across the rubric. The same rubric applies across the shortlist. The overall score summarizes product fit; a specialized workflow can make a lower-ranked product the better choice. ### All 22 criteria and weights - Decision versus raw signal 10 percent - Account-level detection 10 percent - Risk-signal accuracy 8 percent - Bot and automation detection 6 percent - False-positive impact 6 percent - Network and device depth 5 percent - Decision and enforcement 5 percent - Explainability 5 percent - Breadth of signals 4 percent - Real-time decisioning 4 percent - Ease of use 4 percent - Integration speed 4 percent - Price at scale 4 percent - Startup-to-enterprise fit 4 percent - Readable analytics 3 percent - Documentation and SDKs 3 percent - Support quality 3 percent - Pricing transparency 3 percent - Market maturity 3 percent - Feedback loop 2 percent - Data retention 2 percent - Mobile coverage 2 percent ## Interpret close scores A difference of 0.1 on this editorial scale is a close call. It is not a measured accuracy gap. Start with your use case and examine the underlying strengths: a mobile SDK, account linkage, edge mitigation, payment decisions or cost control may outweigh the overall ranking for your product. ## Product facts and pricing Public plan facts are linked to vendor pages in each review. The baseline review was 21 October 2026, with Rupt and Maskbreak plans updated on 23 October 2026. Each review retains its source-check date. Where no comparable public package is confirmed, the table says pricing on request. Billing units differ: identifications, events, calls and fraud checks are not interchangeable. ## Evidence and limitations Vendor pages document their own plans and capabilities. We use those pages as sources for product descriptions and current public access, and distinguish them from our editorial ranking. We do not treat vendor marketing accuracy claims as comparative test results. Buyers should run a representative parallel test before enforcing a new fraud decision. ## Corrections If a plan, feature or source has changed, send the exact URL, quoted passage and current vendor documentation to [editor@protectproduct.com](mailto:editor@protectproduct.com) . The research desk will check the claim against the same rubric and update confirmed facts. Send a dated source so the editorial desk can verify the affected plan or capability. ### Read alongside [Scorecard ↗](/#compare) [Parallel test guide ↗](/guides/parallel-fraud-tool-test/) [Editorial standards ↗](/about/)