Markwell
Risk Intelligence API · LATAM

Mark the risk.
Decide with data.

One endpoint. A score calibrated by country and line of business. Explainable factors on every response. Markwell is the risk intelligence layer for LATAM insurers.

  • · HMAC SHA-256
  • · TLS 1.2+
  • · Law 8968 / CR
  • · p90 < 500ms

Risk score

78

High risk · auto

  • Clean history
  • High-loss region
  • Age bracket
N°01The problem

Underwriting and claims are still decided blind.

Three friction points that erode margin in every policy cycle.

Manual rules that don't scale

Underwriting depends on inherited criteria and the underwriter's intuition. The result: inconsistent decisions and high loss ratio.

Fraud caught after payout

Anomalous patterns surface after the money is gone. Reactive detection erodes margin and reserves.

Portfolio visibility weeks behind

Exposure and loss-ratio reports arrive late. By then, risk concentration is already a fact.

N°02Product

One URL.
The score and the why.

You send the JSON with the applicant's data, signed with HMAC SHA-256. You get a score from 0 to 100, the risk band, and the factors that explain it with their weight. No webhooks. No state. No surprises at audit time.

POSTapi.markwell.ai/v1/score200 OK · 138ms

Request

X-Signature: t=1716070000, v1=a8b9c…

{
  "ramo": "riesgos_trabajo",
  "actividad_ciiu": "4520",
  "provincia": "Puntarenas",
  "edad": 38,
  "num_empleados": 45,
  "suma_asegurada": 15000000,
  "siniestros_previos": 1,
  "antiguedad_cliente": 2,
  "canal": "agente"
}

Response

{
  "score": 97,
  "nivel_riesgo": "ALTO",
  "prima_sugerida": 215000,
  "factores": [
    { 
      "label": "Actividad CIIU de alto riesgo laboral", 
      "impacto": "alto"
    },
    { 
      "label": "Puntarenas — zona de riesgo elevado", 
      "impacto": "alto" 
    },
    { 
      "label": "1 siniestro previo registrado", 
      "impacto": "medio" 
    }
  ],
  "alerta_fraude": false,
  "probabilidad": 0.9722,
  "tokens_restantes": 99991,
  "modelo_version": "2.1.0",
  "roc_auc_modelo": 0.8483,
  "tenant": "Markwell Demo",
  "request_id": "mw_demo_fixed_1778975358.522399"
}
score0–100 · int
bandlow · mid · high · adverse
factorstop variables + SHAP weight
modelversioned id · auditable

One score. Four automated decisions.

Underwriting

Accept, reject, refer.

Score bands → automatic decision on 84% of the pipeline. The remaining 16% reaches the underwriter with context.

Pricing

Per-applicant rate.

Expected loss adjusts the premium continuously. Goodbye to three flat tiers.

Renewal

Retain the good risk.

Re-score the book month over month and focus retention spend where LTV justifies it.

Claims

Fast track for low risk.

Policies with a high score at issuance go through fast-track handling on minor claims.

N°03Impact

Underwrite better.
Pay fewer claims.

Applied to a synthetic portfolio of 10,607 active Workers' Compensation policies, the scoring model identifies the high risk that traditional rules miss. The figures below are projections; real results are calibrated against your historical book.

−8pp

projected loss-ratio improvement

Applied to a synthetic portfolio of 10,607 active Workers' Compensation policies.

+18%

in quote-to-bind

Incremental conversion when the decision on low and mid scores is automated.

−62%

in time to policy

p90 from quote to issuance. The score removes 4 manual steps.

480ms

p90 latency

Regional edge. Designed to add zero friction to your quoting flow.

Synthetic data representative of the Costa Rican market. Projections, not guaranteed results.

N°04Who it's for

Built for every player in the insurance chain.

Same API, different value depending on where you sit in the policy lifecycle.

Insurers

Raise consistency and reduce loss ratio in mass underwriting.

Insurtechs

Ship digital products with scoring from day one without building a data team.

Brokers & agencies

Triage quotes before committing capacity or going to market.

Reinsurers

Analyze concentration and exposure on the books you cede in.

N°05Security & compliance

Built on infrastructure your auditor recognizes.

HMAC signature on every request, encryption in transit, minimum required data, and operation on AWS. No surprises in the security review.

HMAC SHA-256 signature

Every request is signed with a shared secret. The X-Signature header authenticates origin and integrity — and prevents replay with a timestamp.

TLS 1.2+ with ACM

Certificates managed by AWS Certificate Manager. No cleartext traffic between your stack and the API.

API key — OAuth 2.0 on roadmap

Key authentication today, in addition to HMAC. OAuth 2.0 flows for enterprise integration in 2026.

Data minimization

We only ingest the fields needed for scoring. No unnecessary PII.

SOC 2 on roadmap

SOC 2 Type I planned for 2027. Progress reports available under NDA.

AWS us-east-1

Serverless infrastructure on AWS with high availability.

N°06Pricing

Pricing built for every stage of the business.

We talk with you to define volume and SLA. No surprise numbers, no annual minimums.

Starter

Insurtechs and brokers getting started.

  • Per-call pricing
  • Free sandbox
  • Low monthly minimum
  • Self-serve onboarding via docs
Talk to sales →
Most popular

Growth

Mid-size insurers.

  • Monthly subscription with included volume
  • Per-call overage
  • Standard SLA
  • Dedicated support channel
Talk to sales →

Enterprise

Tier-1 insurers and reinsurers.

  • Annual contract with custom volume
  • Guaranteed SLA
  • VPC or on-prem deployment
  • Assisted onboarding
Talk to sales →
N°07Questions

The most common
questions.

Something not here? Email us at mario.perez@markwell.ai.

The first working score takes 3 to 5 days. A production integration with your policy system fits in a standard sprint.

N°08 — Let's talk

Let's talk
about your portfolio.

I'll get back to you within 24 hours. Founder-led, no middlemen.

Mario Pérez

Founder, Markwell

Alajuela, Costa Rica

Book 30 min