The Conan-Holder model: predicting SME failure within 3 years

In brief

Published in 1979 by Joël Conan and Michel Holder, this default-prediction model was built on a sample of French industrial SMEs. It remains widely used in French-speaking Belgium because the balance sheet structure and SME fabric are very similar. While the Altman Z-Score predicts failure within 1 year, Conan-Holder targets a 3-year horizon — the reference indicator for a strategic decision (equity entry, medium-term loan, long-duration mandate).

The formula

1

Z = 0.24·X1 + 0.22·X2 + 0.16·X3 − 0.87·X4 − 0.10·X5

With five ratios:

Note: coefficients on X4 (−0.87) and X5 (−0.10) are negative — the more a company spends on interest or personnel relative to what it produces, the lower the score. The opposite of Altman, where all coefficients are positive.

The five interpretation bands

Conan and Holder published a grid of 3-year failure probability:

The score is deliberately bounded around zero — easy to memorise, easy to explain in a credit committee: "above zero we look, below it we protect ourselves."

Worked example on a Belgian SRL

Take the same Brussels SRL as in our Altman Z'' guide, with a few additional figures needed for Conan-Holder:

Ratios:

• EBITDA ≈ Operating result + Depreciation = 95,000 + 80,000 = 175,000

• X1 = 175,000 / 820,000 = 0.213

• X2 = (380,000 + 280,000) / 1,200,000 = 0.550

• X3 = 750,000 / 1,200,000 = 0.625

• X4 = 18,000 / 2,100,000 = 0.0086

• X5 = 850,000 / 1,250,000 = 0.680

1
2
3

Z = 0.24 × 0.213 + 0.22 × 0.550 + 0.16 × 0.625 − 0.87 × 0.0086 − 0.10 × 0.680
  = 0.0511 + 0.1210 + 0.1000 − 0.0075 − 0.0680
  = 0.1966

Verdict: "very good" band (Z > 0.16) — 3-year failure probability under 10%. Consistent with the Z''-Score of 3.20 on the same company: the two models converge, which substantially reinforces confidence in the diagnosis.

Conan-Holder vs Altman: which one?

Wrong question. The two models answer complementary questions:

Recommendation: compute both and look at the consensus. If both models fall in the same band, the signal is robust. If they diverge, the company is probably atypical (special sector, fast growth, restructuring) and deserves a manual qualitative review. The Company Belgium API exposes the computed consensus in the academicScores block with a confidence level (high / medium / low).

Belgian adaptations

A few attention points when applying Conan-Holder to Belgian annual accounts:

• EBITDA / EBE: the Excédent Brut d'Exploitation is not a standardised rubric in the Belgian chart of accounts. We rebuild it as operating result + depreciation + operating provisions (rubrics 9901 + 630 + 631/4).

• Realisable + Cash: in Belgian accounting, these two notions are not separated from inventory in the abridged schema. We approximate with current assets (rubric 29/58), making X3 slightly optimistic.

• Personnel costs and value added: these rubrics (62, 9800) are available in the full schema but not always in the abridged one. For micro-companies X5 must sometimes be dropped — the term is then set to zero, making the score slightly more favourable. To understand which company files which schema, see our guide on the full schema vs abridged schema.

Typical use cases

• Banks and credit institutions: medium-term loan decisions (3 to 7 years)

• Credit insurance (Atradius, Coface, Euler Hermes): pricing of risk on annual commercial exposure

• Accountants and statutory auditors: assessment of going concern at year-end

• Investment funds: pre-screening of acquisition or equity targets

• CFOs and treasurers: yearly tracking of the main supplier scoreboard

Beyond the score

Conan-Holder remains a statistical filter. Before a final decision, cross it with:

• BCE monitoring to catch weak signals (director changes, address transfer, capital changes)

• UBO verification to identify true ultimate beneficial owners — see our guide on the Belgian UBO register

• A search in the Belgian Official Gazette for recent publications (AGM, conversions, mergers)

• A check against sanctions and PEP lists if you are a profession subject to AML