Let’s be honest—this isn’t just accounting. It’s philosophy wrapped in spreadsheets.
Understanding the IFRS 17 Contract Grouping Framework
The IFRS 17 standard, effective from January 1, 2023, overhauled how insurers report liabilities and profits. No more smoothing. No more hiding mismatched risks. Everything’s out in the open. And at the heart of it? How contracts are grouped. Because if you group them wrong, your entire financial narrative warps. The thing is, contract grouping dictates measurement precision, profit recognition patterns, and even capital allocation decisions.
What Is a Portfolio of Insurance Contracts?
A portfolio is the top-level container. It holds contracts with similar risks and profit sources. Think life insurance in Germany. Or motor policies in Thailand. You can’t mix those—not unless you enjoy explaining to auditors why mortality risk is bundled with flood exposure. The standard allows flexibility, yes, but with teeth: if risks or profit drivers differ materially, they belong in separate portfolios.
And that’s where judgment starts creeping in. Two insurers might sell “term life” products. One lumps all under age 60 into one portfolio. The other splits by underwriting quality—standard vs. substandard. Both could be defensible. But only one might survive a regulator’s deep dive.
What Defines a Cohort?
Within each portfolio, you slice further—into cohorts. These are contracts issued within a 12-month period. January 2023 policies. July 2024 renewals. Each gets its own cohort. Why? Because experience changes. A spike in claims in 2023 shouldn’t drag down 2025 profits if underwriting tightened. Cohorts preserve that timeline integrity.
But—and this is a big but—renewals complicate things. A policy renewed in 2025, even if originally written in 2020, might join the 2025 cohort if it’s renegotiated. Otherwise, it stays put. The issue remains: where do you draw the line on “substantive change”?
The Profitability Test: Where Judgment Reigns
Here’s the pivot point: initial and expected future cash flows must be onerous or profitable as a group. Not per contract. As a whole. If a cluster of policies is expected to lose money, it’s onerous. Must be measured separately. That changes everything.
Imagine selling long-term care insurance in 2018, assuming low claim rates. By 2023, claims surge. Now, that 2018 cohort is underwater. You can’t average it out with a profitable 2022 cohort. The standard forbids it. Which explains why some insurers now carve out legacy blocks even if they share risk profiles.
And here’s what people don’t think about enough: profitability isn’t just about premiums and claims. It includes acquisition costs, investment returns, and even tax effects. One insurer might include hedging gains. Another excludes them. Both follow IFRS 17—but reach different groupings. There’s no single right answer. Just defensible ones.
Because the model hinges on expected cash flows, a 5% shift in discount rates can flip a cohort from profitable to onerous. I am convinced that this sensitivity makes grouping one of the most volatile parts of IFRS 17 implementation—more so than discounting or modeling assumptions.
Timing and Risk Characteristics: The Dual Filter
IFRS 17 doesn’t just say “group by risk.” It demands two filters: risk and timing. Contracts must share both to stay together. A 10-year term policy and a 1-year term policy—even if identical in coverage—might not belong in the same cohort if their cash flow timing diverges enough. That said, “diverges enough” isn’t defined. It’s up to you.
Risk Characteristics: More Than Just Mortality
Mortality, morbidity, lapse rates—yes. But risk also includes behavioral patterns, investment volatility, and even regulatory shifts. A group of policies in Argentina might carry currency risk. The same product in Switzerland doesn’t. Can they be in the same portfolio? Only if the insurer hedges the FX exposure effectively. But hedging isn’t always complete. Hence, segmentation.
To give a sense of scale: one European insurer had to split its “international life” portfolio into 14 sub-portfolios just to isolate sovereign risk in emerging markets. Not because the products differed. But because economic environments did.
Timing of Cash Flows: The Hidden Divider
Even policies with similar risks can’t be grouped if their cash flows are materially out of sync. A front-loaded acquisition cost structure versus a back-loaded claims profile creates mismatch. And that’s exactly where grouping fails silently. You might think you’re simplifying reporting. But you’re distorting results.
For instance, a 2020 cohort with high upfront commissions but low early claims will show different profit emergence than a 2023 cohort with level costs. Merging them smooths volatility. But it violates IFRS 17. So you don’t merge. You segment.
Portfolio vs. Cohort: A Practical Breakdown
Let’s contrast: portfolios are strategic. Cohorts are tactical. A portfolio might cover all U.S. disability insurance. It’s broad. But within it, 57 cohorts could exist—one per issuance month over nearly five years. Each cohort tracks its own loss component, risk adjustment, and contractual service margin.
Now, is this efficient? Not really. But accuracy trumps efficiency here. And honestly, it is unclear whether smaller insurers will sustain this level of granularity without automation. The administrative load is real.
We’re far from it being common practice to merge cohorts post-issuance. Some thought they could, but the standard’s wording is strict. Once split, always split—unless contracts are onerous at inception and later become profitable (or vice versa), triggering a re-evaluation.
Common Pitfalls and Industry Workarounds
The problem is, real-world data is messy. Policies lapse. Riders get added. Reinsurance changes mid-term. And yet, grouping must remain stable. So what do insurers do? Some use “proxy cohorts” for minor modifications. Others apply aggregation thresholds—say, ignoring differences under 2% of total liability.
That’s a gray zone. The standard doesn’t explicitly allow aggregation. But enforcing pure cohort purity could mean thousands of micro-groups. Suffice to say, many are betting that regulators will accept materiality-based simplifications.
One insurer in Singapore reduced 18,000 contract forms into 210 cohorts by applying risk bands and issuance windows. Smart? Maybe. Defensible? They hope so. Because if auditors challenge it, the restatement costs could hit $15 million—just in external fees.
Frequently Asked Questions
Can Two Contracts with Different Premiums Be in the Same Cohort?
Yes—if they share risk and timing profiles. A $100/month and $1,000/month life policy can coexist in a cohort if both are fully underwritten, same term, same demographic band. Premium size alone doesn’t disqualify. But if higher premiums correlate with lower lapse rates (wealthier clients), then risk differs. And that’s where grouping breaks.
What Happens When a Portfolio Becomes Mixed?
If a portfolio starts mixing profitable and onerous cohorts over time, you don’t restructure the portfolio. You track each cohort separately. The portfolio stays. But measurement diverges. The issue remains: financial statements may show stable portfolio-level results while underlying cohorts swing wildly. Transparency improves. Simplicity dies.
Do Reinsurance Contracts Follow the Same Grouping Rules?
No. Reinsurance is under IFRS 17 too, but grouping follows different logic. It’s based on underlying risks assumed, not issuance timing. So reinsurers often have broader cohorts. Which explains why their CSM (contractual service margin) patterns look smoother than direct writers’.
The Bottom Line
Grouping under IFRS 17 isn’t a mechanical checklist. It’s a judgment call wrapped in economics. You can follow the letter of the standard and still get it wrong if your view of risk is outdated. My take? Many insurers are underestimating how much culture—not just systems—needs to change. Actuaries, accountants, and underwriters must debate assumptions constantly.
And here’s a personal recommendation: don’t wait for perfect data. Start grouping now, even with proxies. Because when the audit comes, they won’t care that your system wasn’t ready. They’ll care that your logic was inconsistent.
One final thought: I find this overrated idea that uniformity across firms is possible. With so much room for interpretation, financials won’t be comparable—not really. But maybe that’s the point. Maybe the goal isn’t comparison. Maybe it’s truth in storytelling. (Even if the story is complicated.)
After all, insurance has always been about uncertainty. Now, so is its accounting.