IFRS 17 Reporting Process Gaps: What Breaks in the Close Cycle

Most IFRS 17 build plans I have seen are strong on methodology and strong on accounting-engine selection. The bit that gets underbaked, consistently, is the reporting process itself — the cycle that moves numbers from the actuarial model through the subledger, into the general ledger, onto the disclosure schedules, and into an audit pack that reconciles to all of those. First-close is where that underbake shows up, usually at 11pm on a day the reporting team had not planned to still be at the office.

What follows is the list of process gaps I walk through with clients before we run a dry-run close. Each one is a specific place where the pipeline breaks when it meets real data, real dates, and real reviewers. Fixing them after go-live is expensive. Fixing them during parallel runs is merely annoying.

1. Run cadence mismatch between actuarial and finance

Actuarial model runs used to happen on a monthly or quarterly cadence driven by regulatory reporting. IFRS 17 requires an actuarial run into the subledger every reporting period, on a timeline finance sets — not actuarial. If the actuarial model takes five days to produce a full run and finance needs the subledger posted by day three, the cadence is broken before the cycle starts.

The fix is structural. Either the actuarial run has to accelerate — parallel processing, model simplification, pre-close estimates that get trued up — or the reporting timeline has to absorb the run duration. Most insurers I have worked with end up doing both: a hard-close view produced fast for posting, and a slower refined view produced later for analysis and disclosure.

2. Subledger-to-GL mapping that nobody owns

The IFRS 17 subledger produces journal entries with more dimensions than legacy general ledgers were designed to absorb. Group of contracts, portfolio, cohort, coverage period, measurement model, remaining vs incurred — the subledger wants to post at that level. The GL may only accept two or three of those dimensions natively.

The mapping between them is neither an actuarial job nor an accounting-engine job. It is a finance systems job, and it tends to fall into a gap between the two teams. Who owns the mapping, who approves changes to it, and how it is version-controlled across closes — those questions need named owners. Without them the mapping becomes tribal knowledge held by whoever built the first version, and it breaks silently the first time somebody leaves.

3. The data reconciliation that runs in three directions

Under IFRS 17 the reporting team reconciles actuarial inputs to model outputs, model outputs to subledger postings, and subledger postings to general ledger. Three reconciliations, run every close, each with its own control totals.

Each one needs a defined reconciliation point, a materiality threshold, and a documented break-resolution process. The temptation in early closes is to resolve breaks by adjusting the numbers until they agree. That works once. It does not work when the auditor asks to see the evidence trail behind the adjustments.

The right pattern is to reconcile at natural control points — total gross cash flows, total CSM, total insurance revenue — and to have the break-resolution steps logged in the same workflow that produced the numbers. If the reconciliation lives in a spreadsheet separate from the pipeline, the reconciliation is already in trouble.

4. Disclosure production that lags measurement by a week

IFRS 17 disclosures are long. The analysis of insurance revenue, the movement in CSM, the movement in risk adjustment, the movement in liability for remaining coverage and liability for incurred claims, the reconciliation of insurance contracts by measurement model — these all need to be produced with audit-grade evidence each period.

The common gap is that measurement is completed, the subledger is posted, and the general ledger is closed, and only then does the disclosure team begin building the tables. By then the reporting timeline has no slack and the tables are assembled by hand from multiple sources. First-time disclosure production I have seen takes three to five weeks on that pattern.

The fix is to have disclosure production as a direct consumer of the subledger. Every disclosure table should map to a query against the subledger dimensional model. If the disclosure needs a number that the subledger cannot produce, that is a subledger design gap, not a disclosure gap — fix it upstream, not in the disclosure.

5. The audit workpack that does not exist until the auditor asks

Auditors reviewing IFRS 17 closes want to see a package: the methodology papers, the data lineage, the model run log, the reconciliations, the journal postings with their source, the judgements made during close, the disclosures with their underlying calculations. That package exists in theory. In practice, for the first close, it is usually assembled in the week the auditor arrives from whatever the teams happen to have.

Building the workpack as an output of the close cycle, not as a post-close scramble, is the difference between a two-week audit and a two-month audit. It means every step in the close needs to log its own evidence in a structured form — who ran what, when, against what inputs, with what outputs, and whether the review passed. The audit then reads the log rather than reconstructing it.

6. Judgement capture during close — not afterwards

IFRS 17 closes involve judgement. Coverage-unit choice, risk adjustment calibration, transition methodology application, assumption updates, onerous contract identification — these all involve choices that have to be documented. The documentation is easier to write when the choice is fresh than when the auditor asks about it two months later.

The process gap here is that most close cycles have no designated place to capture judgement. The actuary who chose a coverage unit basis, the finance manager who approved an assumption change, the reviewer who signed off on the risk adjustment quantile — those decisions need a structured record produced at the time the decision is made. A judgement log attached to each close run, with timestamps and approvers, closes this gap cleanly. Retrofit-documenting judgements post-close is the single most common reason first-time IFRS 17 audits extend.

7. The handoff between close cycles

Each reporting period’s close produces outputs that become inputs to the next period — closing CSM balances, locked-in discount rates, cohort assignments, coverage-unit remaining balances, acquisition cash flow run-off schedules. The handoff between close cycles is where silent errors accumulate.

If a cohort is mis-assigned in period one, the error propagates. If a discount curve is not stored with enough precision at period-end, reconstructed CSM in period two drifts. If the closing liability for remaining coverage is posted at a rounded value but carried forward at full precision, a tiny inconsistency is baked into every future period.

The handoff gap closes when period-end outputs are persisted to a single controlled store, versioned, signed off, and read back at period-start as inputs without re-derivation. The principle is simple: what you closed at is what you open with. In practice, it requires explicit storage, access control, and a design that refuses to rebuild period-end numbers on the fly.

Closing thought

The reporting process under IFRS 17 is where methodology, systems, and governance meet. It is also where each of them can quietly undermine the others. Run cadence is a systems-and-planning problem. Mapping ownership is a governance problem. Reconciliation is a methodology-and-data problem. Disclosure production is a systems problem. Audit workpack is a governance problem. Judgement capture is a governance problem. Handoff is a systems problem.

Treating the reporting process as the integration layer for all three — not as a downstream consequence of methodology — is the shift that separates insurers whose first close lands cleanly from insurers whose first close becomes a story told at year-end. The first close is difficult in any case. It does not need to be a story.