TL;DR

IFRS 17 rarely stalls on the accounting. It stalls on five practical things: messy contract data, a CSM nobody can reconcile, assumptions that drift between actuarial and finance, testing that runs out of time, and governance that never named who owns the number. Sort those five and the standard falls into place. This guide covers each one, with the fixes we’ve used on live GCC insurer builds, plus a quick readiness self-check.

Most IFRS 17 projects don’t fail on the standard. They fail on the plumbing. The five things that actually stall an insurer are messy contract data, a Contractual Service Margin nobody can reconcile, assumptions that shift between actuarial and finance, a testing window that runs out, and governance that never decided who owns the number. Solve those five and the accounting takes care of itself. This guide walks each one, with the fixes we’ve used on live GCC implementations.

What makes IFRS 17 so hard to implement?

IFRS 17 is hard because it forces three functions that rarely talk to each other to agree on one set of numbers, every quarter, on a deadline. Actuarial owns the cash flows and risk adjustment. Finance owns the ledger and the CSM roll-forward. IT owns the data pipe feeding both. When any one of them works off a different assumption or a different data cut, the results stop tying out. That’s the whole problem in one sentence.

The measurement rules get the headlines. GMM, VFA, the Premium Allocation Approach, coverage units, discount curves. But teams that already know the standard cold still miss go-live. Why? Because knowing the model and running it on real, dirty portfolio data every close are two very different jobs.

So let’s be specific about where projects actually break.

Why is data readiness the number one IFRS 17 blocker?

Data readiness is the top blocker because IFRS 17 needs data at a grain most legacy systems never stored: contract level, grouped into annual cohorts, split by profitability, tracked from inception. Old policy admin systems held premiums and claims. They rarely held issue dates clean enough to build cohorts, or the granularity to separate onerous contracts from profitable ones.

Here’s what that looks like in practice. A mid-size GCC insurer pulls ten years of motor and medical policies for transition. Issue dates are blank on maybe one in five records. Currency fields mix local and USD with no flag. Reinsurance links to the underlying policy are held in a separate spreadsheet a since-departed analyst maintained. None of that is an IFRS 17 problem. All of it stops IFRS 17 dead.

Fix it early or pay for it forever. The teams that survive spend the first months on data, not modelling:

• Profile the source data before you design anything. Count the blanks. Find out which fields you can actually trust.
• Build the cohort logic on paper first, then in code. Annual cohorts, portfolio, profitability bucket. Get sign-off from actuarial and finance together before a single row moves.
• Set a data dictionary and freeze it. Every downstream disagreement traces back to two people meaning different things by “premium.”

Skip the profiling and you’ll rebuild the model three times as bad records surface at each close. We’ve watched it happen.

Why can’t anyone reconcile the CSM?

The CSM is hard to reconcile because it isn’t a balance you can look up. It’s a running story: recognised at inception, unwound with interest, adjusted for changes in future service, then released to profit over coverage units. Every quarter it moves for four or five separate reasons, and the roll-forward has to explain each one. Miss a driver and the closing balance won’t tie, and your auditor will find it.

The Contractual Service Margin is where finance and actuarial collide hardest. Actuarial produces the cash flow movements. Finance has to turn those into a CSM roll-forward the ledger accepts. If the two teams book the same change in different periods, or one nets an adjustment the other grosses up, the number drifts. Small at first. Ugly by year end.

What works is a shared roll-forward template both teams sign off on, line by line, before close. Opening CSM, interest accretion, new business, changes in estimates, release for the period, closing CSM. One template. One owner per line. No “we’ll reconcile it later,” because later never comes cleaner than now.

How do IFRS 17 actuarial assumptions cause profit volatility?

Assumptions cause volatility because IFRS 17 splits how their changes hit the accounts, and a small move in the discount rate or lapse rate can swing results in ways management didn’t see coming. Changes tied to future service adjust the CSM and stay off the income statement. Changes tied to past or current service flow straight to profit. Get the split wrong and earnings jump for reasons the board can’t explain.

The discount rate is the sharpest example. IFRS 17 lets you build it top-down or bottom-up, and the risk adjustment for non-financial risk sits on top of that. Two reasonable actuaries can land on curves that differ by 30 to 40 basis points. On long-tail medical or life business, that gap moves the liability by real money. So the method has to be set once, documented, and defended, not rechosen each quarter because the number looked off.

One honest limit here. We can build you a stable assumption framework, but no framework removes volatility that’s genuinely in the book. If your medical portfolio reprices hard every year, IFRS 17 will show it. The job is to make sure the volatility you report is real, not an artefact of a method nobody agreed on. That distinction is what earns trust with the audit committee.

For the mechanics of setting and governing these inputs, our note on IFRS 17 actuarial assumptions goes a level deeper on discount curves and risk adjustment.

Why does testing always run out of time?

Testing runs out of time because teams treat it as the last phase instead of a parallel one, then discover during dry-run closes that the data, the model, and the ledger disagree. By the time those breaks surface, go-live is weeks away and there’s no runway to fix root causes. So they patch symptoms, and the patches follow them into production.

A real IFRS 17 close touches a long chain: source data, cohort assignment, cash flow projection, CSM roll-forward, disclosures, ledger posting. Each handoff can break. You need several full dry-run closes on real data to find where, and each one takes weeks to run and unpick. Start that in the final quarter and the maths simply doesn’t work.

Book the dry runs into the plan from the start. Three end-to-end closes on production-grade data, spaced out, each with a fixed list of what “passed” means. Then compare the IFRS 17 result against your old basis and force yourself to explain every material difference. If you can’t explain a movement, you don’t understand your own model yet. Better to learn that in a dry run than in front of the auditor.

Who actually owns the IFRS 17 number?

This is the one nobody wants to answer, and it’s the one that sinks projects. Ask an insurer who signs off the closing CSM and you often get a pause. Actuarial thinks finance. Finance thinks actuarial. The CFO assumes someone told them. IFRS 17 lives exactly on that seam, so an undefined owner means an unowned number.

Governance failures don’t show up as errors. They show up as delay. A reconciliation break sits in an inbox for two weeks because it’s not clearly anyone’s job. A methodology question stalls because no forum exists to decide it. Multiply that across a close and a 20-day timetable becomes 45.

Fix it with three things, and only these three at first. A named owner for each output. A standing forum, monthly, where actuarial and finance resolve methodology together and log the decision. And an escalation path to the CFO for anything unresolved in a set number of days. Boring? Yes. It’s also the difference between a close that lands and one that drifts. For the wider control environment around this, our internal audit and governance team sets these frameworks up as standard.

IFRS 17 implementation readiness: a quick self-check

Before you plan the next phase, score yourself honestly against these. If you’re shaky on three or more, your timeline is at risk, whatever the project plan says.

Five questions. If any answer is “I’d have to check,” that’s your next priority. Where does your team actually sit on the governance one?

What does a realistic IFRS 17 timeline look like?

A realistic build runs 12 to 18 months for a mid-size GCC insurer, and the split surprises people. Roughly the first third goes to data and design, the middle third to modelling and ledger integration, and the final third to dry-run closes and fixing what they surface. If your plan front-loads modelling and leaves testing till the end, flip it.

Insurers already live and reporting have their own next problem: the close is too slow and too manual. That’s a different engagement, and honestly a harder one, because you’re changing an engine while it runs. If that’s you, the fix is process and automation, not another model rebuild.

Frequently asked questions

What is the biggest challenge in IFRS 17 implementation?

Data readiness is the biggest single blocker. IFRS 17 needs contract-level data grouped into annual cohorts by profitability, a grain most legacy policy systems never captured. Blank issue dates, mixed currencies, and broken reinsurance links stop the project before modelling even starts.

Why is the CSM so hard to reconcile under IFRS 17?

The Contractual Service Margin isn’t a fixed balance. It moves each period for several reasons: interest accretion, new business, changes in estimates, and release for coverage. Actuarial produces the movements and finance books them, so mismatched timing between the two teams makes the closing balance drift.

How long does IFRS 17 implementation take?

For a mid-size GCC insurer, a full build runs 12 to 18 months. Plan for roughly a third on data and design, a third on modelling and ledger integration, and a final third on dry-run closes and remediation. Front-loading modelling and leaving testing last is the most common way to miss go-live.

Which measurement model should insurers use under IFRS 17?

Most short-duration general insurance qualifies for the Premium Allocation Approach, which is simpler. Life and long-term contracts use the General Measurement Model, and contracts with direct participation features use the Variable Fee Approach. The eligibility test for the PAA is where teams most often go wrong.

Do we need new software to comply with IFRS 17?

Usually yes for the calculation engine and CSM roll-forward, since spreadsheets rarely hold up to audit at scale. But the tool is the smaller decision. A capable engine on ungoverned data still produces numbers nobody can reconcile, so fix data and ownership first, then choose the software.

Where to start if you’re behind

Start with the two cheapest fixes that unblock everything else: name who owns the CSM, and profile your source data. Neither needs a budget line. Both surface the problems you’d otherwise hit at go-live, when they cost ten times more to fix. Do those this month.

If you want a second set of eyes on where your build actually stands, that’s exactly what our actuarial team does. Send us your timeline and we’ll tell you, straight, which of the five is most likely to bite. Book a free IFRS 17 readiness review and bring your hardest question.