Loss Triangle Data Preparation: Pricing-Ready Triangles
On this page
- Key Takeaways
- What Is Loss Triangle Data Preparation?
- Which Claims Listing Fields Feed a Loss Triangle?
- How Is a Loss Triangle Built From a Claims Listing?
- How Can One Data Error Distort Development Factors?
- Which Data Preparation Errors Break Loss Triangles?
- Which Checks Should Pricing Teams Run Before Using a Triangle?
- How Does Insurnest Support Loss Triangle Data Preparation?
- What Should a Chief Actuary Ask Cedents for at the Next Renewal?
- Key Terms
- Sources
- Frequently Asked Questions
Last reviewed: November 2026
The cedent's claims file arrives as three spreadsheets, each with a different valuation date and its own column names.
The pricing deadline is next week, and the actuary spends the first three days just finding out which claims are missing.
Loss triangle data preparation is the work of turning a cedent's claims listing into a development triangle that pricing actuaries can trust.
It works when each field is mapped to a role, the origin basis is chosen on purpose, and the triangle is tested against known errors before any factor is selected.
Key Takeaways
- The listing decides the triangle. Errors in the data become errors in every factor.
- Cumulative and incremental must reconcile. If they do not, the build is wrong.
- The origin basis is a choice. Accident, underwriting and report year answer different questions.
- Missing closed claims understate development. Ask for full listings, not open-claim reports.
- Thresholds create false development. Note every reporting threshold and when it changed.
What Is Loss Triangle Data Preparation?
Loss triangle data preparation is the step between receiving cedent claims data and selecting development factors. It is data work, not modelling.
It feeds experience rating such as burning cost reinsurance pricing, and sits inside the actuarial pricing request workflow.
The Casualty Actuarial Society's text by Friedland, Estimating Unpaid Claims Using Basic Techniques, sets out the data checks used below.
Which Claims Listing Fields Feed a Loss Triangle?
The claims listing fields that feed a loss triangle each play one role: key, origin, age, value or filter.
| Listing field | Triangle role | Treatment |
|---|---|---|
| Claim number | Key | Must stay stable across valuations; check for duplicates |
| Date of loss | Origin (accident year) | Required for losses occurring analysis |
| Policy inception or underwriting year | Origin (underwriting year) | Required for policies attaching analysis |
| Report date | Origin (report year) and lag | Shows late reporting |
| Valuation date | Development age | Age = months from start of origin year to valuation |
| Paid to date | Cumulative paid value | Confirm it is cumulative, not period paid |
| Outstanding reserve | Added to paid for incurred | Confirm case reserve only, no IBNR |
| Currency | Conversion | Convert at one fixed rate across all valuations |
| Status | Filter and check | Open, closed or reopened |
| Recoveries and expenses | Value adjustment | State gross or net, with or without expenses |
| Event code | Filter | Separate cat losses from attritional development |
How Is a Loss Triangle Built From a Claims Listing?
A loss triangle is built from a claims listing by assigning each claim an origin year and a development age, then summing values in each cell.
The example below is illustrative, in USD thousands, with year-end valuations from 2023 to 2025.
What does the illustrative claims listing contain?
The illustrative claims listing contains six claims, with cumulative paid amounts at each 31 December valuation.
| Claim | Date of loss | Reported | Paid at 2023 | Paid at 2024 | Paid at 2025 |
|---|---|---|---|---|---|
| C1 | March 2023 | 2023 | 100 | 250 | 300 |
| C2 | August 2023 | 2023 | 50 | 150 (closed) | 150 |
| C3 | November 2023 | 2024 | 0 | 80 | 200 |
| C4 | February 2024 | 2024 | n/a | 120 | 260 |
| C5 | September 2024 | 2024 | n/a | 40 | 90 |
| C6 | May 2025 | 2025 | n/a | n/a | 70 |
How are cumulative paid amounts placed in the triangle?
Cumulative paid amounts are placed in the triangle by summing all claims from one accident year at each development age.
| Accident year | 12 months | 24 months | 36 months |
|---|---|---|---|
| 2023 | 150 (100 + 50 + 0) | 480 (250 + 150 + 80) | 650 (300 + 150 + 200) |
| 2024 | 160 (120 + 40) | 350 (260 + 90) | |
| 2025 | 70 |
The 24-to-36 month factor for 2023 is 1.354 (650 ÷ 480). The volume-weighted 12-to-24 factor is 2.677 ((480 + 350) ÷ (150 + 160)).
How do incremental values differ from cumulative values in the triangle?
Incremental values differ from cumulative values in the triangle because they show only what was paid in each period.
| Accident year | 0–12 months | 12–24 months | 24–36 months | Check |
|---|---|---|---|---|
| 2023 | 150 | 330 (480 − 150) | 170 (650 − 480) | 150 + 330 + 170 = 650 |
| 2024 | 160 | 190 (350 − 160) | 160 + 190 = 350 | |
| 2025 | 70 | 70 |
Each row of incremental values must add back to the latest cumulative value. If it does not, a claim has been lost or double counted.
How does the origin basis change the same triangle?
The origin basis changes the same triangle by moving claims between rows.
C3 occurred in November 2023 but was reported in 2024. On a report year basis it moves to the 2024 row, so 2023's 24-month value falls to 400.
Friedland notes that underwriting year data is frequently used by reinsurers.
Under that basis, each claim's row follows its policy's inception year, not its loss date.
How Can One Data Error Distort Development Factors?
One data error can distort development factors when a closed claim drops out of a later listing.
Suppose the cedent's 2025 file is an open-claims report. C2, closed in 2024 at 150, is missing from it.
The 2023 row at 36 months then shows 500 (300 + 200), not 650. The 24-to-36 factor falls from 1.354 to 1.042 (500 ÷ 480).
Applied to 2024's 350 at 24 months, the projection at 36 months falls from about 474 (350 × 1.354) to about 365 (350 × 1.042).
That is roughly 109 of development lost from one missing claim. Nothing in the file looks wrong unless claim counts are checked.
Which Data Preparation Errors Break Loss Triangles?
The data preparation errors that break loss triangles mostly come from how the cedent's listing was produced.
| Error | Symptom in the triangle | Test | Fix |
|---|---|---|---|
| Closed claims dropped | Development flattens in older years | Claim counts by origin fall between valuations | Ask for a full listing including closed claims |
| Period paid read as cumulative | Values fall between ages | Any negative incremental paid | Confirm field definition and rebuild |
| Reopened claims given new numbers | Claim counts jump, old claim stays closed | Same insured and date of loss twice | Link the reopened claim to the original |
| Currency at each valuation's rate | Development moves with exchange rates | Recompute at one fixed rate | Convert all valuations at one rate |
| Reporting threshold changed | Sudden growth at one valuation | Compare minimum claim size by year | Apply one threshold to all years |
| Cat losses mixed with attritional | Spiky factors in event years | Event codes on large claims | Separate cat claims, using a large loss register |
| Origin set by report date | Late claims shift rows | Date of loss vs report date | Rebuild on the intended basis |
| Claim numbers changed after migration | Claims vanish and reappear | Match on date of loss and amount | Map old numbers to new |
Thresholds matter most for excess of loss listings. If a cedent lists claims above 50% of the retention, claims can enter the listing only as they grow.
Which Checks Should Pricing Teams Run Before Using a Triangle?
The checks pricing teams should run before using a triangle follow the data review components set out by Friedland.
- Consistency with financial statements. Reconcile paid and case reserves to the cedent's reported figures.
- Consistency with prior data. Compare with last year's submission and explain any restated values.
- Data reasonableness. Investigate questionable values, such as large negative payments.
- Data definitions. Confirm what each field means rather than relying on its label.
- Claim count tie-out. Counts by origin should not fall between valuations unless the cedent explains why.
How Does Insurnest Support Loss Triangle Data Preparation?
Insurnest supports the intake step, before the triangle is built. It does not build triangles or select factors.
Reinsurance Email Intelligence reads broker emails with their claims files, PDFs and spreadsheets together.
It extracts claims history, loss ratios, gross and ceded premium and exposure into one structured submission view, and flags missing or incomplete information.
Mapping to origin and age, the error tests above and all actuarial judgement stay with your pricing team.
What Should a Chief Actuary Ask Cedents for at the Next Renewal?
A chief actuary should ask cedents for claims data in a form that makes the error table above unnecessary.
- Full listings at each valuation. Include closed and nil claims, not only open ones.
- Both paid and case reserve fields. Each confirmed as cumulative to date.
- Date of loss, report date and policy inception. So any origin basis can be built.
- Stated threshold and currency basis. With the date of any change.
- Event codes on large claims. So cat losses can be separated.
Gaps in these fields belong on the broker query list early, as covered in incomplete reinsurance submissions.
Key Terms
- Origin year: The year a claim is grouped under, by loss date, inception or report date.
- Development age: Months from the start of the origin year to the valuation date.
- Cumulative value: Total paid or incurred to date for an origin year.
- Incremental value: The movement in one development period.
- Development factor: The ratio of cumulative values at two consecutive ages.
Sources
- Casualty Actuarial Society: Estimating Unpaid Claims Using Basic Techniques, J. Friedland (Version 3, July 2010)
- Insurnest: Reinsurance Email Intelligence
Frequently Asked Questions
What is loss triangle data preparation?
Loss triangle data preparation is the work of turning a cedent's claims listing into a development triangle that pricing actuaries can trust. It covers mapping fields, choosing the origin basis, building cumulative and incremental values by development age, and testing the data for gaps and errors.
What is the difference between a cumulative and an incremental loss triangle?
A cumulative loss triangle shows the total paid or incurred for each origin year up to each valuation date. An incremental loss triangle shows only the movement in each period. Cumulative values are the running sum of incremental values, so both triangles must reconcile.
Which origin basis should a reinsurance loss triangle use?
A reinsurance loss triangle should use the origin basis that matches the treaty and the pricing question. Accident year suits losses occurring covers, underwriting year suits policies attaching covers and is common among reinsurers, and report year helps when reporting lags are a concern.
How do you build a loss triangle from a claims listing?
To build a loss triangle from a claims listing, assign each claim an origin year from the date of loss or inception, calculate its development age at each valuation date, sum cumulative paid or incurred by origin and age, then derive incremental values and reconcile both.
Why do closed claims disappear from cedent claims listings?
Closed claims disappear from cedent claims listings when the cedent sends an open-claims report instead of a full listing. Paid amounts on closed claims then drop out of later valuations, which makes development look lighter and understates projected losses.
What data checks should be run before using a loss triangle for pricing?
Before using a loss triangle for pricing, check that totals reconcile to the cedent's financial data, that the data is consistent with last year's submission, that values look reasonable, and that every field definition is confirmed. Then test claim counts, currencies, thresholds and large losses.
How do reporting thresholds affect reinsurance loss triangles?
Reporting thresholds affect reinsurance loss triangles because claims below the threshold are missing from the listing. If the threshold changes between years, or claims cross it as they develop, the triangle shows growth that comes from the threshold, not from the claims.
Should loss triangles use paid or incurred claims?
Loss triangles for reinsurance pricing are usually prepared on both paid and incurred bases. Paid claims avoid reserving judgement but develop slowly, while incurred claims include case reserves and show emerging losses sooner. Comparing the two helps reveal changes in reserving practice.

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 →