Skip to content
pulsaf5.com · all systems nominal
status EN sign in
pulsaf5.com · all systems nominal 99.987% uptime · trailing 12 months

Decision Intelligence Platform Product

Score, resolve, route, and act — one real‑time signal for every fintech decision.

Pulsaf5 turns fragmented transaction, identity, and behavioral data into a single performance signal, so risk, credit, and growth teams ship decisions 9x faster without trading away accuracy.

Book a 20-minute demo See the platform
  • 01Risk Scoring
  • 02Identity Resolution
  • 03Behavioral Signals
  • 04Routing & Orchestration
Schematic SVG of a transaction timeline with annotated risk-score markers along a 24-hour axis
4 composable surfaces on one signal layer
11 days median model deployment
4.2B anonymized transactions since 2019
99.987% trailing‑12‑month uptime

02 Product surfaces

One platform, four composable surfaces — engineered for every stage of the decision pipeline.

Risk Scoring, Identity Resolution, Behavioral Signals, and Routing & Orchestration share a single signal layer and a unified API, so each surface inherits the lineage, calibration, and governance of the one before it.

01 · Flagship

Risk Scoring

Real‑time risk and credit decisions on every transaction. Models trained on 4.2 billion anonymized transactions across 41 fintech verticals; adaptive AML calibration; deployed in production at challenger banks and high‑growth processors.

Latency p95
38 ms
False‑positive reduction
23% in pilot deployments
Coverage
Transaction, credit, AML
Risk Scoring on the platform →

02

Identity Resolution

Deterministic and probabilistic matching across device, account, and biometric graphs — built for KYC‑grade audit trails and instant re‑verification under load.

Match precision
99.4% at 10⁻⁴ FPR
Identity graph
Cross‑device, cross‑entity
Audit trail
Lineage‑preserving
Identity Resolution on the platform →

03

Behavioral Signals

Velocity, session, and lifecycle features distilled into a per‑entity signal stream — the same feature store that powers Risk Scoring, exposed for your growth and credit models.

Feature freshness
Sub‑second
Signal types
Velocity, session, lifecycle
Re‑use
Across risk, credit, growth
Behavioral Signals on the platform →

04

Routing & Orchestration

Policy‑driven decisioning that routes a single transaction across processors, KYC vendors, and internal workflows — with full replay and shadow‑mode evaluation.

Connectors
140+ pre‑built
Modes
Live, shadow, replay
Policy
Versioned, diff‑able
Routing & Orchestration on the platform →

03 Technical diligence

Built on a corpus, a team, and a certification posture your security review will recognize.

A short, evidence‑driven note for the diligence checklist — covering training data scale, deployment cadence, the certifications we already hold, and the depth of the research bench behind every model shipped.

4.2B anonymized transactions trained on since 2019, across 41 fintech verticals
11 days median model deployment, against a Gartner 2024 industry benchmark of 14 weeks
SOC 2 II · ISO 27001 · PCI‑DSS L1 full audit reports available under NDA in 24 hours
38 research staff, including 7 PhDs in quantitative finance from Stanford, MIT, and ETH Zürich
1T+ risk‑scoring events processed, with the 1 trillionth event logged in Q2 2024
380+ fintechs production users across 22 countries — the basis for our 2024 NPS of 74

What your security review will actually look at

The same five questions come up in nearly every procurement cycle. We publish the answers here so the technical evaluation can run in parallel with the commercial one.

  1. Where do the models come from? A 38‑person research team — 7 PhDs in quantitative finance among them — operating against a corpus of 4.2 billion anonymized transactions spanning 41 verticals. Model lineage, training data snapshots, and evaluation reports are versioned and exportable.
  2. How long does deployment actually take? Eleven days median, against a Gartner 2024 benchmark of fourteen weeks. The work that compresses the timeline is upstream — connectors, identity graphs, and feature stores are already provisioned; your team plugs in policy and risk appetite.
  3. What certifications do you hold? SOC 2 Type II, ISO 27001, and PCI‑DSS Level 1. Full audit reports are available under NDA within 24 hours, and our Information Security team will join the technical review on request.
  4. What happens to our data? Customer data is isolated at the tenant level, encrypted at rest and in transit, and never contributes to shared training corpora. Data residency options cover the United States and the European Union today, with additional regions on the public roadmap.
  5. How do we measure ROI? Customer‑reported average ROI of 5.8x within the first 12 months of deployment, measured against the customer's pre‑Pulsaf5 baseline. We will co‑design the baseline with your team before the first transaction is scored.

For a deeper dive, the fourth edition of the State of Fintech Risk report documents the benchmarks behind these numbers. A PDF is shared on request from your Pulsaf5 contact.

Surface 01 — Risk Scoring flagship surface · v4.2 model
Schematic transaction timeline showing five events with risk-score annotations and a 0.70 auto-decline threshold.
Fig. 01 — A representative 12-second scoring window. Nodes above 0.70 auto-route to review or decline.

Risk Scoring, built on 4.2 billion transactions and the documented 23% false-positive reduction.

Pulsaf5's Risk Scoring surface is the editorial centerpiece of the platform — a real-time decision engine trained across 41 fintech verticals since 2019, returning calibrated probability and reason codes in a single API call. It is the layer our customers reach for first, and the layer our research team iterates on weekly.

Training corpus
4.2B anonymized transactions across 41 verticals (consumer, SMB, cross-border, neobank, issuing, marketplace).
Inputs accepted
transaction.amount, transaction.currency, transaction.merchant_category, device.fingerprint, session.ip_risk, identity.email_age_days, behavioral.session_depth, plus 38 derived features.
Output schema
{ score: float[0–1], band: enum(clear|review|decline), reason_codes: string[≤6], model_version: semver, latency_ms: int }
Latency target
p50 38ms · p95 92ms · p99 184ms (trailing 30 days, all customers).
Deployment window
Median 11 days from contract signature to first production decision, against the Gartner 2024 industry benchmark of 14 weeks.
  • 23%reduction in false-positive fraud alerts across pilot deployments at three top-10 neobanks.
  • 5.8×customer-reported average ROI within the first 12 months of deployment.
  • 99.987%platform uptime across the trailing twelve months — decision paths remain available.
Surfaces 02 & 03 — Identity & Behavior

Identity Resolution and Behavioral Signals — first-class inputs, not afterthought connectors.

Scoring a transaction in isolation is a 2017 pattern. Pulsaf5 ingests identity, device, and session context as schema-equal signals alongside the transaction itself, so the model sees the same person across onboarding, login, and checkout — and the same person across two devices, two emails, and three addresses.

  1. 02

    Identity Resolution

    Stitch fragmented KYC, account, and device records into a single canonical entity — before a score is requested.

    Inputs
    email, phone_e164, national_id_hash, device.fingerprint, cookie_id, account_id, mailing_address (normalized).
    Outputs
    entity_id (stable ULID), confidence_score, linked_identifiers[], first_seen_at, last_seen_at, cluster_graph_ref.
    Latency target
    p95 110ms for resolve · p95 240ms for full cluster graph.
    • KYC stitching at onboarding, reducing duplicate review queues by collapsing alias records.
    • Returning-customer detection across new devices, without relying on cookies alone.
    • Sanctions and adverse-media graph joins, returned as linkable flags on the same entity.
  2. 03

    Behavioral Signals

    Session, device, and interaction telemetry — collected via a 9KB JS snippet or server-side events — folded into the same scoring payload.

    Inputs
    session.depth, session.typing_cadence, session.touch_event_count, device.model, device.os_build, network.tls_fingerprint, behavioral.mule_proxy_score.
    Outputs
    behavior_vector[64-dim], bot_probability, mule_proxy_probability, session_risk_band.
    Latency target
    Stream-evaluated; available on every /score call without additional round-trips.
    • Device intelligence surfaced as reason codes ("device_emu_detected", "headless_browser").
    • Session signals distinguish a hurried human from a credential-stuffing bot at the same IP.
    • Behavioral vectors are stored per-entity, so a returning attacker in week 6 looks different from week 1.

Integration notes for technical diligence

  • Single REST endpoint POST /v1/score accepts identity and behavioral payloads inline — no separate calls required.
  • OpenAPI 3.1 schema published under /platform; type-safe clients available for TypeScript, Python, Go, and Ruby.
  • Backwards compatibility guaranteed for two minor versions; deprecations announced 90 days in advance.
  • SOC 2 Type II, ISO 27001, and PCI-DSS Level 1 certified — full audit reports available under NDA within 24 hours.
Surface 04 — Routing & Orchestration the action layer · write-back decisions
Connector graph showing Pulsaf5 at the center with 14 representative pre-built connectors to core banking, KYC, and data warehouse systems.
Fig. 02 — A representative slice of the 140+ pre-built connectors. Pulsaf5 writes decisions back, not just reads signals in.

Routing & Orchestration closes the loop — score to decision to downstream system, in one transaction.

Most analytics layers stop at the score. Pulsaf5's Routing & Orchestration surface writes the decision back into payments, issuing, onboarding, and data warehouse systems in real time — so risk, credit, and growth teams stop copy-pasting reason codes into five different consoles.

Decision primitives
route, hold, release, step_up_3ds, manual_review, decline, write_to_warehouse, emit_webhook.
Routing rules
Declared in YAML or the dashboard; evaluated in <40ms; supports AND/OR nesting, time-of-day overrides, and per-entity exceptions.
Write-back destinations
Core banking ledgers, issuing BINs, KYC platforms, case management tools, Snowflake / BigQuery / Redshift / Databricks.
Failure handling
At-least-once delivery with idempotency keys; dead-letter queue surfaced in the dashboard; replay API for ops teams.

140+ pre-built connectors — the categories that matter for diligence

  • Core banking Plaid, Finicity, MX, Yodlee, direct cores via ISO 20022.
  • KYC / AML Alloy, Persona, Sum&Substance, Onfido, Veriff, traditional bureau pulls.
  • Data warehouse Snowflake, BigQuery, Redshift, Databricks, Postgres, S3 (Parquet).
  • Payments rails Issuing processors, 3DS providers, RTP/FedNow endpoints, card networks (acquirer-side).
  • Case & ops Salesforce, Zendesk, internal ticketing via webhook with signed payloads.
Book a 20-minute demo Read the orchestration reference →