For fintech compliance + onboarding pipelines
PEP + sanctions screening for fintech onboarding
Fintech onboarding (a partner, an introducer, an AR, or a customer in some product surfaces) carries hard regulatory liability + a soft brand-risk piece. The screening cost has to be sized for both onboarding (one-off) and ongoing monitoring (high-frequency). Per-call pricing at the sanctions vendor's retail rate is unaffordable at fintech scale, and the sub-second latency budget on a customer-onboarding flow rules out subscription tools that gate at 5-10s of API latency. Stratum runs sub-100ms p95 on warm matchers + £0.69 entry rate matches Veriphy.
What the screen returns
Each call to /v1/sanctions/screen returns sanctions matches (OFSI / UN / EU lists with score + sanction regime), PEP matches (FATF Recommendation 12 categories: current / former-within-12mo / family / close-associate; HIGH / MEDIUM / LOW risk tier), and adverse-media mentions grouped by severity. List-version timestamps in every response keep the audit trail FATF-compliant. The shape stays stable across releases (additive only).
Latency that fits the onboarding flow
Sub-100ms p95 on warm matchers. The indexes load on Lambda cold-start (~150-300ms one-off); warm path is in-memory. Add ~30ms for the API gateway. For a customer-onboarding flow that already takes 8-12s end-to-end (form + KYC + biometric + AML), the screening adds ~130ms — invisible inside the existing latency budget. For real-time payment-flow screening, the same shape holds.
OpenSanctions Match API option
For the most demanding workflows, we also expose a direct OpenSanctions Match API path (€0.10 per query, retail surfaces only — never daily monitoring per our cost guard rail). This delivers the freshest possible PEP catalogue + lower false-positive rate via OpenSanctions' own confidence scoring. Recommended for high-stakes single-name screens (corporate-onboarding partners) rather than transaction-level checks. Set `USE_OPENSANCTIONS=true` on the suite Lambda env.
Frequently asked
What latency should I budget for?
Sub-100ms p95 on warm matchers. Cold-start adds ~150-300ms one-off (the index loads). For a payment-flow that needs sub-50ms, talk to us about Pro for dedicated capacity + warmup.
How do you handle fuzzy matching?
The matcher handles Cyrillic-Roman, German umlauts, Arabic transliteration. Threshold defaults to 0.7; pass `?threshold=0.85` for stricter matching at higher precision. Pin behaviour with the spot-check tests against known-real names (Erdoğan, Gaddafi, Al-Assad).
How fresh is the PEP data?
The OpenSanctions PEP catalogue refreshes weekly via the Sunday-morning cron; list-version timestamp in every response confirms which version your check ran against. For Match API path: live (always reads the current OpenSanctions version).
Continuous monitoring at fintech scale?
The continuous-monitoring tier (£19.99/mo for 100 entities; £49/mo for 500) re-screens daily + emails a digest. For 5000+ monitored entities — talk to us about a custom tier.
GDPR + UK GDPR posture?
We're UK-hosted (AWS eu-west-2) + UK GDPR clean by default. No data leaves UK. Anthropic's Claude (used in our sales chatbot) is configured with UK-data residency + zero-retention through the API.