The Data Quality Problem in Actuarial Practice
Every actuarial estimate — whether a pension liability valuation, a general insurance reserve, or an IFRS 17 CSM calculation — begins with data. Policy records, claims histories, membership files, exposure data: these inputs are the raw material from which actuaries construct their models. When that raw material is compromised — by missing records, inconsistent coding, duplicated entries, or outdated information — the resulting estimate carries an uncertainty that no amount of modelling sophistication can eliminate.
In the East African context, data quality challenges are pervasive. Legacy policy administration systems that were not designed with actuarial modelling in mind, manual data entry processes that introduce transcription errors, and historical records that were not digitised systematically all contribute to a data environment that actuaries must navigate carefully. The first step is acknowledging that data quality is a material actuarial risk — one that belongs in the risk management framework alongside model risk and parameter uncertainty.
What a Data Governance Framework Provides
A data governance framework establishes the policies, processes, and accountabilities needed to ensure that data meets the quality standards required for actuarial use. At its core, it answers three questions: Who is responsible for data quality? What standards must data meet? How are quality issues identified and resolved?
For insurance companies and pension funds, effective data governance requires engagement across multiple functions — IT, operations, finance, and the actuarial team. The actuarial function's role is to define the specific data requirements for valuation, pricing, and reserving models; to communicate those requirements clearly to data owners; and to validate that requirements are being met. This is an active role, not a passive one.
“A technically perfect model built on poor data produces a precise answer to the wrong question. In regulated financial services, that is not a technical problem — it is a governance failure.”
— Niloyd Associates Data & Analytics Practice
Conducting an Actuarial Data Audit
An actuarial data audit is a structured review of the data used in actuarial models, assessing completeness (are all required fields populated?), accuracy (do the data values reflect the underlying reality?), consistency (are data in different systems aligned?), and timeliness (is the data current enough for the purpose?). The audit produces a data quality report that identifies gaps, quantifies their materiality, and recommends remediation steps.
For pension schemes undergoing an actuarial valuation, a data audit ahead of the valuation date allows data issues to be corrected before they propagate into the liability estimate. For general insurers, a data audit as part of the IBNR reserving process ensures that claims development patterns are not distorted by incomplete notification records. For life companies transitioning to IFRS 17, a data audit is essentially a prerequisite — the standard's disclosure requirements cannot be met without clean, granular policy-level data.
Strengthening Predictive Modelling with Better Data
The intersection of data quality and predictive analytics is increasingly important as insurers and pension funds adopt more sophisticated models for pricing, risk selection, and experience monitoring. Machine learning and statistical models are particularly sensitive to training data quality — biased or incomplete training data produces models that perform poorly on new data, and may embed systematic errors that are difficult to detect without rigorous validation.
Model validation that focuses only on methodology and parameterisation — without assessing the quality of the underlying data — provides incomplete assurance. Niloyd Associates advocates for a validation framework that includes explicit data quality assessment: reviewing the source, preparation, and transformation of data before it reaches the model, and establishing ongoing monitoring to detect data quality degradation over time.
Key insights
A good model on bad data produces confident errors
Sophisticated actuarial models cannot compensate for incomplete, inconsistent, or inaccurate input data. Precision without accuracy is not a virtue — it obscures the true uncertainty in the estimate.
Data governance is an actuarial responsibility
The actuarial function must be an active participant in data governance — not merely a consumer of whatever data is available. Identifying data requirements, defining data standards, and validating inputs are core actuarial duties.
Audit trails protect the actuary as much as the institution
A documented, reproducible data preparation process — from source systems to model inputs — is essential for regulatory scrutiny, peer review, and the defence of actuarial judgements.
Conclusion
Actuarial credibility depends on more than technical skill. It depends on the quality of the data on which technical judgements are built. Institutions that invest in data governance, conduct regular actuarial data audits, and maintain clean, well-documented data environments produce more reliable estimates, withstand regulatory scrutiny more effectively, and make better-informed decisions. Niloyd Associates provides actuarial data audit services, data governance framework development, and model validation support — helping regulated financial institutions build the data foundations that actuarial practice demands.
Work with us
Have a question about this topic? Our advisors are available to discuss specific mandates and regulatory questions.
