Most consulting case studies read like victory laps. Ours include the difficult parts and the things we would do differently. We publish them because the work is more useful that way.
Two anonymised engagements below.
Case study one — IFRS 17 ETL and reporting build for a large SA life insurer
Context. A large South African life insurer running roughly thirty sub-funds across legacy SAP and AS400 source systems. The IFRS 17 implementation programme had delivered the calculation engine but left the data lineage between source policy administration and disclosed CSM movement reconstructed each quarter by hand. The audit findings on the prior cycle flagged it.
The problem. Migrate the sub-fund close onto a defensible IFRS 17 lineage. Reconcile historical balance build-up across the legacy sources, so a quarter-end number could be traced from source transaction to disclosed figure without a reconstruction step.
What we did. Built the ETL layer between SAP, AS400 and the IFRS 17 calculation engine as versioned, lineage-tracked pipelines. Reconstructed the historical balance build-up across both legacy estates to a defensible starting position. Industrialised the CSM movement and waterfall production. Parallel-ran for two cycles against the existing process before cutover. Wrote the methodology notes and lineage documentation to the standard the statutory audit team had asked for.
What was difficult. The historical balance reconciliation. The two legacy systems did not agree on transaction date for a subset of historical movements, and the differences had been absorbed into manual overlays over years. We had to redo the reconciliation more than once, and explain to the audit team why the new build did not produce the same number as the prior cycle for a known set of cases — the new number was right, the prior number was the overlay. That conversation took longer than the technical work.
What we would do differently. Start the audit conversation in week one, not week ten. The overlay reconciliation surprise would have been easier to land if the audit team had been part of the diagnostic phase from the start.
What the client got. Quarterly close completed two days earlier per cycle. The first audit on the new build passed without methodology findings. CSM movement and waterfall reproduce six months later, from the same source data, bit-for-bit.
Case study two — BI estate replacement for a SA private-wealth business
Context. A South African private-wealth business running a legacy reporting suite that had grown by accretion over fifteen years. Spreadsheet and dashboard estates were duplicating numbers across functions, and a planned platform decommissioning forced the question of what to replace it with.
The problem. Decommission the legacy reporting suite and replace it with a governed, business-owned BI estate. Prove the new estate matched the legacy at decimal precision before cutover.
What we did. Defined the executive KPI catalogue from scratch with the business unit owners. Built the new data products as the single source for each KPI. Migrated the dashboards in tranches, by business unit. Parallel-ran the new estate against the legacy for six months, reconciling every KPI to decimal precision and resolving the differences (mostly: legacy was using a stale lookup table; sometimes: legacy was right and we had a transformation bug). Retired the legacy on the planned date.
What was difficult. The parallel-run was longer than scoped. Two business units pushed back on the new KPI definitions late in the programme, and the resolution required reopening the catalogue work. We absorbed the cost rather than passing it on, but the lesson was: lock the KPI catalogue with all business units before starting the build, not during it.
What we would do differently. Run the KPI catalogue lock as a separate, fixed-fee diagnostic phase before the build is scoped, not as the first week of build. The business units treat the conversation more seriously when it is the only thing on the table.
What the client got. Legacy reporting suite retired on plan. Ongoing reporting cost halved. One executive number per KPI, traceable end-to-end, owned by the business unit that produces it. The new estate has survived a leadership change since.
More case-study briefs
Shorter case-study briefs covering specific engagements. Downloadable as PDF.
CASE STUDY
General insurance pricing
Pricing methodology rebuilt around domain-aware data and reproducible test runs.
Download briefCASE STUDY
Predictive modelling for profitability
Predictive models applied to portfolio profitability, with lineage from data to decision.
Download briefCASE STUDY
Orchestrating an IFRS 17 workflow
End-to-end orchestration of a quarterly IFRS 17 reporting cycle.
Download briefCASE STUDY
Data management under IFRS 17
Data architecture and lineage for an insurer running IFRS 17 reporting at scale.
Download briefCASE STUDY
Funeral product pricing
Funeral pricing methodology and operational design for a SA life insurer.
Download briefCASE STUDY
Open-source financial modelling platform
An open-source modelling stack deployed in production, with version control and reproducibility.
Download briefCASE STUDY
Explaining neural networks
Explainability methods applied to neural-network credit and underwriting models.
Download briefCASE STUDY
Modernising actuarial data management
Actuarial data management modernisation, end-to-end.
Download briefCASE STUDY
Transforming actuarial assumption management
Assumption-management transformation across a large actuarial estate.
Download brief