An executive guide to insurance underwriting data governance in South Africa, with a focus on rating factors, exposure data, pricing confidence, and reinsurance readiness.
Many South African insurers do not have a pricing sophistication problem. They have an underwriting data control problem.
The board sees pressure on claims ratios, lapse experience, reinsurance costs, capital strain, and new business margins. Management responds with pricing reviews, new rating models, portfolio actions, and distribution changes. Yet the data feeding those decisions is often inconsistent, incomplete, or poorly owned.
For short-term insurers, exposure data may not reliably describe the risk being insured: property construction type, security measures, vehicle use, driver profile, insured values, flood exposure, or business interruption limits. For life insurers, occupation, income band, smoker status, medical loadings, policy changes, and beneficiary information may be captured differently across channels and administration platforms.
This is why insurance underwriting data governance South Africa is an executive issue, not an actuarial housekeeping exercise. Pricing can only be as reliable as the rating factors, exposure records, underwriting decisions, and claims feedback that support it.
Insurers tend to examine pricing models when loss experience deteriorates. That is understandable, but it can miss the earlier failure point: the quality of data captured at quote, policy inception, endorsement, renewal, and claim.
A motor book may appear under-priced for a certain province. On review, the issue may not be the territorial factor itself. The problem may be that garaging address, regular driver, vehicle use, and tracking device status are captured inconsistently by brokers, call centres, digital channels, and legacy systems. A household book may show poor fire experience in one segment, but construction type and roof material may be blank for a material portion of the portfolio.
In life insurance, a mortality or morbidity review can be distorted when occupation categories are too broad, income information is stale, or underwriting exclusions are stored in free text rather than structured fields. The result is not merely untidy reporting. It affects product profitability, fairness of pricing, claims expectations, and the credibility of management action.
Executives should therefore ask a simple question before approving a major pricing change: are we correcting the risk, or are we compensating for weak data?
Rating factors are the characteristics used to assess and price risk. In short-term insurance they may include age of driver, vehicle type, property location, sum insured, alarm status, business activity, construction class, prior claims, or cover options. In life insurance they may include age, sex, smoker status, health indicators, occupation, income, education, policy term, benefit type, and underwriting decision.
The governance issue is not whether these factors appear in a pricing manual. It is whether the insurer can show who owns them, how they are defined, where they are captured, how changes are approved, and how exceptions are monitored.
Consider a commercial property insurer. “Security level” may mean one thing to underwriting, another to a broker portal, and another to claims assessors after a burglary. If that factor is used in pricing but not validated consistently, the insurer may grant discounts that are not supported by the actual risk. Over time, the portfolio appears to perform worse than expected, but the root cause sits in field-level governance.
Rating factor governance should cover four practical controls: approved definitions, mandatory capture rules where material, validation at source, and regular comparison against claims outcomes. This does not require a large bureaucracy. It requires named accountability between underwriting, actuarial, claims, finance, compliance, and technology.
Exposure data describes what the insurer is actually on risk for. It is central to pricing, accumulation monitoring, reinsurance, reserving, capital modelling, and board reporting.
For a short-term insurer, exposure data includes insured locations, sums insured, limits, deductibles, occupancy, fleet size, geographic concentration, and policy period. For a life insurer, it includes sum assured, benefits, lives covered, policy duration, underwriting class, premium pattern, and policyholder characteristics.
The South African context makes this harder. Many insurers still run multiple policy administration systems following acquisitions, product launches, broker schemes, or legacy migrations. Load-shedding and operational disruption can also create workarounds, delayed updates, and batch processing gaps. A spreadsheet maintained by a specialist team may become the “real” exposure view, while the source system remains incomplete.
That creates risk at executive level. A CFO may sign off on financial results using one exposure basis, while the actuarial team uses another for experience analysis and the reinsurance team submits a different bordereau. In a benign period the differences may be tolerated. After a catastrophe, mortality shock, cyber incident, or large commercial fire, they become difficult to defend.
Strong exposure governance means reconciled values, controlled adjustments, traceable changes, and clear agreement on which dataset is used for which decision.
Reinsurers do not only price the claims history. They assess the insurer’s ability to understand and manage its book.
Poor exposure data can weaken reinsurance negotiations. If a property portfolio cannot provide reliable geocoding, construction details, sums insured, occupancy, and accumulation views, reinsurers will price uncertainty into the treaty or impose tighter terms. If a life portfolio cannot explain underwriting classes, policy movements, claims development, and sums at risk clearly, the insurer may face more challenge during renewal.
This matters in South Africa where catastrophe exposure, infrastructure risk, economic pressure, and changing policyholder behaviour all influence risk appetite. Reinsurance capacity is not guaranteed on favourable terms. The insurer that can produce clean, consistent exposure data has a stronger negotiating position than one that relies on manual extracts assembled late in the renewal process.
Boards should treat reinsurance data preparation as a year-round governance discipline, not an annual scramble. The same data used for treaty renewal should connect to pricing, reserving, risk appetite, and portfolio steering.
Underwriting relies on personal and sometimes sensitive information. Life insurers may process health, occupation, lifestyle, financial, and beneficiary data. Short-term insurers may process identity information, addresses, vehicle records, claims history, criminal incident details, and financial information.
POPIA requires lawful processing, purpose limitation, appropriate security, access control, retention discipline, and accountability. In practical terms, underwriting data governance must answer: why do we collect this field, who may use it, how long do we keep it, how is it corrected, and how is it protected?
There is also a conduct dimension. If rating factors are poorly governed, customers may be priced inconsistently or treated unfairly. A broker channel may capture risk information one way, while a direct digital journey applies stricter validation. A renewal model may use stale customer information. A claims decision may reveal that a discount was granted on information never properly verified.
The issue is not only regulatory compliance. It is trust. Insurers asking customers for detailed personal information must be able to show that the data is necessary, controlled, and used responsibly.
Executive committees do not need to manage data fields line by line. They do need a clear view of control over the data that drives underwriting and pricing decisions.
A practical governance review should answer:
This work should sit within the insurer’s broader data strategy, not apart from it. Zorinthia’s insurance data advisory perspective is set out at data strategy for insurance, with practical sector examples available under insurance examples.
Insurers should not begin with a platform decision. They should begin with a portfolio decision.
Choose one material book — for example motor, commercial property, funeral, credit life, disability income, or high-net-worth household — and test whether the rating factors, exposure data, underwriting outcomes, claims feedback, and reinsurance extracts can be reconciled into one trusted view.
If they cannot, the next executive question is not “Which tool do we buy?” It is: which underwriting and pricing decisions are we currently making with data we cannot fully defend?