Treaty Data for Reserving Segmentation: Keys Compared
On this page
- Key Takeaways
- What Is Treaty Data for Reserving Segmentation?
- Which Segmentation Keys Should Reserving Actuaries Compare for Reinsurance Treaties?
- What Decision Rule Decides Whether to Split a Reserving Segment?
- How Do Mixed Attachment Bases Distort a Reinsurance Development Triangle?
- How Should Treaty Teams Prepare Treaty Data for Reserving Segmentation?
- How Does Insurnest Support Treaty Data for Reserving Segmentation?
- Key Terms
- Sources
- Frequently Asked Questions
Last reviewed: November 2026
The latest underwriting year looks remarkably light at twelve months. The reserving committee wonders if the portfolio got better.
It did not. Most new treaties moved from losses occurring to risks attaching, and the triangle never knew.
Treaty data for reserving segmentation is the set of treaty attributes actuaries use to group reinsurance contracts into reserving classes.
It matters because development patterns only work when every cell in a triangle behaves alike, and treaty terms decide that behaviour.
This guide gives you a segmentation key comparison matrix, a split decision rule and a worked example of mixed bases distorting development.
Key Takeaways
- Basis before line. Attachment and year basis change when losses arrive.
- Split for behaviour. Group treaties that develop alike, not those that look alike.
- Guard credibility. Too many small segments give unstable factors.
- Source from the treaty. Basis and type must come from bound terms.
What Is Treaty Data for Reserving Segmentation?
Treaty data for reserving segmentation is the treaty-level attribute set that places each contract in a reserving class.
The claims data itself is prepared separately; see loss triangle data preparation. This guide covers the treaty keys.
Which Segmentation Keys Should Reserving Actuaries Compare for Reinsurance Treaties?
Reserving actuaries should compare seven segmentation keys for reinsurance treaties, each by how it changes development.
| Key | Typical values | What it changes | Default action | Treaty data error |
|---|---|---|---|---|
| Attachment basis | Losses occurring (LOD), risks attaching (RAD) | Period over which losses arrive | Always split | Basis missing, assumed LOD |
| Year basis | Underwriting year (UY), accident year (AY) | Cohort definition | Always one basis per triangle | AY and UY rows mixed |
| Treaty type | Quota share, surplus, per-risk XoL, cat XoL | Reporting speed and severity | Split | Type taken from programme name |
| Line of business | Property, casualty, marine, motor | Tail length | Split | Multi-line treaty in one class |
| Cat and non-cat | Named events vs attritional | Volatility and timing | Split cat out | No event code on claims |
| Layer position | Working vs high layer | Frequency at the layer | Split if credible | Attachment not stored |
| Currency | USD, EUR, GBP etc. | FX movement in development | Convert at fixed rates | Mixed currencies summed |
The CAS text by Friedland sets the goal: "sufficiently homogeneous groupings" without compromising credibility.
Why do losses occurring and risks attaching treaties develop differently?
Losses occurring and risks attaching treaties develop differently because they cover different loss periods for the same treaty year.
A losses occurring 2023 treaty covers losses in 2023.
A risks attaching 2023 treaty covers policies incepting in 2023, so losses can occur until late 2024 on annual policies.
Should reinsurance reserving use underwriting year or accident year?
Reinsurance reserving often uses underwriting year, and Friedland notes that reinsurers often organise claims data that way.
The HMRC General Insurance Manual glossary describes cohorts as claims grouped by accident year, report year or underwriting year.
Pick one basis per triangle and state it in the triangle title.
What Decision Rule Decides Whether to Split a Reserving Segment?
The decision rule that decides whether to split a reserving segment ranks keys by what they change.
| If the key changes… | Then | Example |
|---|---|---|
| When losses can occur | Always split | LOD vs RAD |
| How cohorts are defined | Never mix | UY vs AY |
| Reporting speed or tail | Split if each part stays credible | QS vs XoL; property vs casualty |
| Volatility only | Split out, reserve separately | Named cat events |
| Measurement only | Convert, do not split | Currency at fixed rates |
If a split leaves a segment too thin, merge it with the nearest behaving segment and document why.
How should multi-line reinsurance treaties be split for reserving?
Multi-line reinsurance treaties should be split for reserving with a documented allocation key, not left in a "multi-line" class.
- Premium by class. Use bordereau or cedent premium split by line for each year.
- Claims by class. Use the line on each claim record where the cedent reports it.
- Fallback. Where claims lack a line, allocate in proportion to premium and flag it.
Keep the allocation key with the extract. A key that changes silently between valuations moves development without any claim changing.
How Do Mixed Attachment Bases Distort a Reinsurance Development Triangle?
Mixed attachment bases distort a reinsurance development triangle when the LOD and RAD mix shifts between years. Figures are illustrative.
| Basis | Reported % of ultimate at 12 months | 12-month factor to ultimate |
|---|---|---|
| LOD | 45% | 2.222 |
| RAD | 15% | 6.667 |
| Blended, old mix (80% LOD / 20% RAD) | 39% | 2.564 |
UY 2024 shifted to 20% LOD and 80% RAD, with a true ultimate of 100m: LOD 20m and RAD 80m (USD).
- Reported at 12 months. LOD 20m × 45% = 9m; RAD 80m × 15% = 12m; total 21m.
- Wrong method. 21m × 2.564 (old blended factor) = 53.8m ultimate.
- Right method. 9m × 2.222 + 12m × 6.667 = 20m + 80m = 100m.
The blended triangle understates ultimate by 46.2m, or 46%. The "light" year was a basis shift, not better business.
That error flows straight into IBNR allocation to reinsurers and into reserve comparisons with cedents.
How Should Treaty Teams Prepare Treaty Data for Reserving Segmentation?
Treaty teams should prepare treaty data for reserving segmentation as one row per treaty, layer and year, drawn from bound terms.
- Supply the keys. Type, attachment basis, inception, expiry, class, cat flag, attachment, limit and currency.
- Test completeness. No blank basis; no "multi-line" class without a split rule.
- Freeze at valuation. Keep the extract version each valuation used.
- Flag changes. A treaty whose basis changed at renewal starts a new segment row.
- Reconcile. Treaty count and premium must match the treaty system.
Where cedent and reinsurer views differ, see cedent reserves vs reinsurer reserves.
How do IFRS 17 groups relate to reserving segments?
IFRS 17 groups relate to reserving segments loosely, and the two should be mapped rather than forced to match.
IFRS 17 divides contracts into groups for recognition and measurement. It also bars grouping contracts issued more than one year apart.
Reserving segments follow development behaviour. Keep a mapping table from segment to IFRS 17 group, refreshed each valuation.
How Does Insurnest Support Treaty Data for Reserving Segmentation?
Reinsurance Treaty Management keeps a controlled record per treaty: structure, terms and conditions, participation share, retention and limits.
Changes become controlled treaty versions with an audit trail, so a valuation extract can show which terms applied.
It does not build triangles, estimate IBNR or produce reserving dashboards. Actuaries still own segmentation choices and test the data.
Key Terms
- Losses occurring (LOD): Treaty covering losses that occur during the treaty period.
- Risks attaching (RAD): Treaty covering losses on policies that incept during the treaty period.
- Segmentation key: A treaty attribute used to place contracts in a reserving class.
Sources
- Casualty Actuarial Society: Estimating Unpaid Claims Using Basic Techniques (Friedland)
- HMRC: General Insurance Manual GIM99999 glossary
- IFRS Foundation: IFRS 17 Insurance Contracts
- Insurnest: Reinsurance Treaty Management
Frequently Asked Questions
What is treaty data for reserving segmentation?
Treaty data for reserving segmentation is the set of treaty attributes actuaries use to group contracts into reserving classes, such as treaty type, attachment basis, year basis, currency and line.
Which segmentation keys should reserving actuaries compare for reinsurance treaties?
Reserving actuaries should compare treaty type, attachment basis, year basis, line of business, cat versus non-cat, layer position and currency as segmentation keys for reinsurance treaties.
Why do losses occurring and risks attaching treaties develop differently?
Losses occurring and risks attaching treaties develop differently because a risks attaching year picks up losses on policies written in the year, so losses arrive over roughly two years.
Should reinsurance reserving use underwriting year or accident year?
Reinsurance reserving often uses underwriting year, which matches premium to losses, but accident year suits losses occurring business. Never mix both bases in one triangle.
What decision rule decides whether to split a reserving segment?
The decision rule for splitting a reserving segment is: always split on keys that change the loss period, split on keys that change reporting speed if each part stays credible.
How do mixed attachment bases distort a reinsurance development triangle?
Mixed attachment bases distort a reinsurance development triangle when the mix shifts. In the illustration, a blended pattern projects 53.8m against a true ultimate of 100m.
How should treaty teams prepare treaty data for reserving segmentation?
Treaty teams should prepare treaty data for reserving segmentation by supplying, per treaty and year, the type, attachment basis, inception date, class, cat flag, layer and currency from the bound record.
How do IFRS 17 groups relate to reserving segments?
IFRS 17 groups relate to reserving segments loosely: IFRS 17 divides contracts into groups for measurement, while reserving segments follow development behaviour, so map one to the other.

Hitul Mistry
CEO, Insurnest
An InsurTech leader with more than a decade of experience across insurance and technology, focused on solving business problems with the help of technology. Has worked with brokers, insurance carriers, and reinsurance firms across the India, UAE, and US markets.
View LinkedIn profile →