Actuarial Science · Data Science · CIMA Zone · 14 Countries
The analytics infrastructure for African health insurance
We combine robust statistics, data science and actuarial expertise to help health insurers detect fraud, manage their claims ratio and optimize provider networks.
Diagnosis
Why existing tools fail in Africa
Diagnosis
- International tools are calibrated for mature markets: different fee schedules, variable data quality, CIMA-specific contracting structures they cannot handle
- Control teams manage unmanageable volumes without automatic prioritisation — every file gets the same level of attention
- No shared data infrastructure exists in the CIMA zone: no common reference, no sector benchmark
- Network fraud (provider/insured collusion) is invisible to file-by-file analysis
The Karelytics approach
- Statistical models calibrated on CIMA portfolio specifics: local fee schedules, country-specific claims distributions
- Explainable risk score on every file — each alert is defensible in one sentence before any audit
- Progressive construction of a pooled data asset: the more insurers join, the sharper the detection
- Graph analysis to detect collusion and fraud networks invisible to standard file-by-file controls
Our approach
Data science and actuarial science applied to African health insurance
We apply techniques documented in the scientific literature — anomaly detection, robust statistics, graph analysis — rigorously calibrated for the specifics of CIMA portfolios. Our value is not the invention of the science, but its rigorous application to a domain where it is under-exploited.
Pillar 1
Non-life actuarial science
Domain expertise and deep understanding of African health portfolios — CIMA fee schedules, benefit structures, claims profiles by country and product line.
Pillar 2
Robust statistics & data science
Anomaly detection on incomplete and heterogeneous data — modelling normal behaviour, multidimensional risk scoring adapted to low-volume markets.
Pillar 3
Unsupervised machine learning
Identification of anomaly combinations no human rule would formulate — organised fraud networks, emerging behaviours, provider/insured collusion.
What we detect
Three detection perimeters
Our engine analyses every file across three complementary risk axes. The exact patterns detected are presented during private demonstrations.
Provider irregularities
- Tariff anomalies relative to CIMA fee schedules
- Act / pathology / protocol inconsistencies
- Statistically atypical behaviour relative to peers
- Abnormal concentrations on specific file profiles
Member irregularities
- Inconsistencies in consumption history
- Identity and eligibility anomalies
- Statistically aberrant care-seeking behaviour
- Weak signals of documentary fraud
Network fraud
- Abnormal statistical links between providers and members
- Network structures revealing implicit coordination
- Collective behaviours undetectable file by file
- Details shared during private demonstration
Auditabilité totale — Every alert comes with an auditable statistical explanation. Our engine prioritises — your teams decide.
Long-term vision
Building the African Optum
The ambition: becoming the reference analytics infrastructure for health insurance in francophone Africa.
As our platform onboards more insurers, the richness of the pooled data increases value for everyone — cross-portfolio fraud detection, precise sector benchmarking, emergence of a shared reference framework that exists nowhere in the CIMA zone today.
Our method
A transparent process. No black box.
Data maturity audit
No commitmentAssessment of your available data sources, identification of blind spots, synthetic report delivered within 5 business days.
Personalised feasibility note
Proposed scope, detailed technical approach, measurable expected results and success conditions.
Proof of concept — 30 days
30 daysOn anonymised data, our engine produces its first alerts. Concrete, measurable results at D+30.
Pilot — 6 months
6 monthsDeployment on a defined scope, iterative adjustments, impact measurement on the loss ratio.
Full deployment
Integration into your existing control processes. Team training. Continuous monitoring.
Notre engagement
“Every alert produced by our engine is explainable in one sentence to a claims manager. We do not deliver an opaque algorithm — we deliver a rigorous, auditable system.”
Zéro déploiement sans résultats mesurables prouvés lors du PoC.
Window of opportunity
Why now?
Three convergences make 2025–2026 the unique window of opportunity for Karelytics.
Regulation has just unlocked the market.
The CIMA adopted the digital TPA regulatory framework in 2025. Before this date, delegation of claims management to a digital third-party operator was legally ambiguous across most of the 14 member countries. That barrier has just been removed. Insurers can now legally outsource — and the most agile are actively looking for a partner.
The data asset doesn't exist yet — but the window is closing.
The structured, longitudinal claims dataset that Karelytics is building exists nowhere in the CIMA zone today. In 3 to 5 years, better-capitalised players — local or international — will enter this market. The value of the data asset is at its maximum right now. Every month of delay is a month of lost data and reduced lead.
The market is structurally under-equipped as it must grow.
Insurance penetration in sub-Saharan Africa is 3% of GDP. Governments are pushing hard towards universal health coverage — CMU in Côte d'Ivoire, CMU/CSU in Senegal, CSU in Cameroon. This expansion will multiply the volume of claims to be processed. Without analytics infrastructure, this growth will be unmanageable.
The first-mover position is being built today. Every claim processed is one more data point in the asset.
Business model
Three revenue streams. One underlying asset.
Per-claim fee
For every claim processed and analysed, Karelytics receives an operational fee from the insurer.
Analytics subscription
Insurers subscribe to real-time claims dashboards and management tools.
Data benchmarks
Pooled sector analyses and predictive models sold to institutions and employers.
A dataset that grows with every claim processed. A value that appreciates over time.
Contact
Managing health risk for your policyholders?
Request a data maturity audit or a platform demonstration.