Segmenting receivables for a provision matrix

A single provision matrix assumes every customer behaves like the average. When some customers are much riskier than others, and the mix shifts, that assumption fails quietly. This guide explains how to choose segments, how many to have, and shows how a blended matrix understates the allowance when the mix changes.

By Mirza Fahad Baig, Chartered Accountant. Reviewed by Hamza Fida, Chartered Accountant. 3 minute read.

Short answer

Provision matrix segmentation means building separate loss rates for groups of customers with different credit risk, as IFRS 9 requires when it says a provision matrix should reflect shared credit risk characteristics. Common segments are customer type, industry, geography, size, product and the existence of credit insurance or collateral. Segmenting matters most when the customer mix changes: in this guide's example, one blended matrix would understate the allowance by 4,449 after the share of riskier independent shops grows.

At a glance

IFRS 9 basis
Shared credit risk characteristics
Common segments
Customer type, industry, geography
Also
Credit insurance, collateral, product
Too few segments
Mix changes distort the rates
Too many segments
Too little data in each
Excel
Loss rate builder
Segmenting receivables for a provision matrixIFRS 9 basis: Shared credit risk characteristics; Common segments: Customer type, industry, geography; Also: Credit insurance, collateral, product; Too few segments: Mix changes distort the rates; Too many segments: Too little data in each; Excel: Loss rate builder.KEY FACTS AT A GLANCESegmenting receivables for a provision matrixIFRS 9 basisShared credit riskcharacteristicsCommon segmentsCustomer type, industry,geographyAlsoCredit insurance,collateral, productToo few segmentsMix changes distort theratesToo many segmentsToo little data in eachExcelLoss rate builderTax BakersSegmenting receivables for a provision matrixIFRS 9 basis: Shared credit risk characteristics; Common segments: Customer type, industry, geography; Also: Credit insurance, collateral, product; Too few segments: Mix changes distort the rates; Too many segments: Too little data in each; Excel: Loss rate builder.KEY FACTS AT A GLANCESegmenting receivables for aprovision matrixIFRS 9 basisShared credit risk characteristicsCommon segmentsCustomer type, industry, geographyAlsoCredit insurance, collateral, productToo few segmentsMix changes distort the ratesToo many segmentsToo little data in eachExcelLoss rate builderTax Bakers
Key facts at a glance, as set out in this guide.

Provision matrix segmentation: how should receivables be grouped?

IFRS 9 allows a provision matrix as a practical expedient, but expects the loss rates to reflect groups of receivables that share credit risk characteristics. Useful dividing lines are the ones where loss experience genuinely differs:

  • Customer type: large retailers, independent shops, government bodies, individuals.
  • Industry or geography: customers in a sector or region that moves differently in a downturn.
  • Size and credit rating: a few large, rated customers may be better assessed individually.
  • Credit enhancements: receivables covered by credit insurance or letters of credit lose less.
  • Product or channel: online sales to consumers behave differently from trade credit to businesses.

An example: why a blended matrix fails when the mix changes

Loss rates by segmentLoss rates by segmentCurrent31-60 daysOver 90 daysRetail chains0.2%3%35%Independentshops1.2%12%65%Blended(historical mix)0.5%7.0%51.5%
Independent shops lose far more at every age.

A distributor sells to retail chains and to independent shops. Historically, chains made up most of the receivables, but the company has grown its independent shop business, which now holds 458,000 of the 1,180,000 total, up from a much smaller share.

Allowance
Retail chains, own loss rates4,900
Independent shops, own loss rates24,810
Segmented total29,710
One blended matrix, rates reflecting the historical mix25,261
Understatement from blending4,449

The blended rates were right for the old mix but are too low now, because more of the balance sits with riskier customers. The Segments sheet of the Loss rate builder (Excel) runs this comparison.

How many segments are enough?

Enough to separate the groups whose loss experience clearly differs, but not so many that each segment has too few write-offs to give reliable rates. A segment with one or two write-offs a year will produce volatile, unreliable loss rates. Most companies end up with two to five segments, sometimes with a few large customers assessed individually, and review whether the split still works each year.

When should a customer be assessed individually?

When it is large enough to matter on its own, has its own credit rating, or shows specific signs of trouble, such as missed payments, a restructuring or news of financial difficulty. A customer assessed individually is removed from the matrix, so its balance is not counted twice, and its specific provision is documented separately.

What data is needed to segment?

A customer master file that records each customer's type, industry and region consistently, so that ageing reports and write-off histories can be split the same way. Many companies find that their customer data is not clean enough at first, and segmentation starts with tidying it up.

How is credit insurance reflected?

Receivables covered by credit insurance that is part of the trade terms, or by letters of credit, have lower losses. They can form their own segment with lower loss rates, or the matrix can be applied to the uninsured part only. Insurance bought separately is usually a separate asset, not part of the matrix calculation itself.

How often should segments be reviewed?

At least annually, as part of preparing the year-end allowance, and whenever the business changes: a new customer type, a new market, or a change in credit policy. Testing whether loss rates differ significantly between candidate segments is a good discipline before adding or merging any segments.

See historical loss rates, forward-looking adjustments and the IFRS 9 provision matrix.

Need help applying the standards?

Our Chartered Accountants help finance teams and students apply IFRS and US GAAP to real transactions.

Questions people ask

Why segment a provision matrix?

Because customers with different credit risk have different loss rates, and a single blended matrix becomes wrong when the customer mix changes.

How do you choose segments for a provision matrix?

By shared credit risk characteristics where loss experience differs: customer type, industry, geography, size, credit insurance or product.

How many segments should a provision matrix have?

Enough to separate groups with clearly different losses, but each needs enough data for reliable rates; two to five is common.

When should a customer be taken out of the matrix?

When it is large, rated, or shows specific signs of financial difficulty; it is then assessed individually.

Sources

Every fee, date and rule on this page was taken from these official and primary sources.

  1. IFRS Foundation: IFRS 9 Financial Instruments

Rules and fees change. If you are reading this long after October 4, 2026, confirm the figures with the source before you rely on them.

More in ECL

This guide is general information. It is not tax or legal advice for your situation.