ESG matters to credit analysis when a change in the environment can be connected to a borrower’s ability to repay. Calling a company green or risky is too coarse. I want to understand which operations produce income, what drives costs, where collateral is located and how quickly a change could affect cash flow. This creates a practical connection between banking, reliable information and the systems that maintain it.
I approach this as an analyst seeking to deepen my knowledge of credit risk. The discussion does not replace specialist assessment of a borrower or legal interpretation of applicable rules. My focus is whether assumptions are supported, missing information is recognised and a broad topic can be translated into a question that someone can investigate. Collecting many fields without a clear decision purpose would not be my starting point.
A real case: Crédit Agricole, France, February 2026
On 13 February 2026, the ECB announced EUR 7,551,050 in periodic penalty payments against Crédit Agricole for failing to meet a climate-related and environmental risk materiality-assessment requirement. Its announcement identified 75 full days of non-compliance in 2024. This was a decision concerning that institution, not a general daily tariff automatically applicable to all banks.
I do not infer the quality of individual loans from this announcement. The useful analytical question is whether an institution can demonstrate how it identified material risks. A statement that work is under way does not explain coverage, evidence or exclusions. The process below is my proposed approach to an assessment record, not a reconstruction of how Crédit Agricole organised its internal programme or an assessment of its current position.
Identify the mechanism rather than a label
I would first distinguish physical disruption from changes in business conditions. Flooding might damage a plant or interrupt transport. Higher energy costs might erode margins without damaging any property. Both channels can affect one borrower. A single coloured score is less helpful if it obscures which mechanism matters, when the impact would arise and which assumptions connect the exposure to the borrower’s finances.
For a corporate group I would establish the relationship between the borrower, production facilities and any guarantor. A registered office may be far from the factory. A sector average may help screening but should not silently become a measurement of one company. Missing information should remain identifiable. If an estimate substitutes for it, the record needs the source, reason and limitations so that its apparent precision does not conceal uncertainty.
An illustrative energy-intensive manufacturer
Imagine annual operating receipts of CZK 120 million and operating payments of CZK 100 million, including CZK 20 million for energy. Assume annual debt service of CZK 12 million and, for simplicity, cash flow before further investment. The CZK 20 million surplus covers debt service approximately 1.67 times. These figures are hypothetical and do not come from a borrower file or the regulatory case above.
If energy costs rise by half and selling prices cannot increase, annual payments increase by CZK 10 million. The remaining surplus falls below assumed debt service. This simplified calculation determines neither a rating nor a lending decision. It identifies the next questions: what price contracts exist, what savings are feasible, which liquidity resources are available and how soon would the pressure become relevant to scheduled payments?
Now add an investment in more efficient production. I would not immediately replace future costs with the expected lower amount. I would examine financing, the implementation timetable, production downtime and sensitivity to delay. Savings starting in two years do not necessarily address repayments due in the next six months. I would compare the base case with delayed implementation and a combined scenario involving expensive energy and weaker demand.
Connect assumptions with evidence
I would list the most important assumptions and attach a supporting record to each. Energy consumption could be reconciled with invoices and production volumes rather than relying solely on management estimates. Price protection needs a coverage period, proportion and counterparty. Property data needs a unique identifier and the actual location. Conflicting values should first be documented as discrepancies; their cause should not be assumed before examining the relevant records.
A data team could help connect borrowers, facilities, loans and collateral into a reproducible view. A specialist would confirm business meaning and scenario relevance. My contribution would be to test completeness of those relationships, record exceptions and communicate limitations clearly. Automation is useful once the meaning of a field is established. A faster calculation built on an ambiguous definition is not a more reliable basis for a decision.
Test whether information changes behaviour
A completed form would not be enough evidence for me. I would select cases and investigate whether an identified risk led to an appropriate question, monitoring action or decision. The result does not always need to be more restrictive terms. A reasoned conclusion that a risk is immaterial can also be appropriate. The essential distinction is between a deliberate assessment and an omission disguised by a complete-looking record.
I would also test a material change: a plant moves location, a technology changes or a contract expires. Does the next assessment use the new fact? Who captures the update, and which downstream reports receive it? Following a documented change through the process provides stronger evidence of how the system works than inspecting its latest form alone. It also reveals where ownership becomes unclear between business and IT teams.
For an initial review I would select different information conditions: a large exposure using estimates, a smaller borrower with a specific warning and a case assessed as immaterial. This purposive sample would test process behaviour, not estimate a portfolio-wide failure rate. I would record why each case was chosen and keep an information gap distinct from evidence of repayment difficulty. An incomplete record warrants a qualified conclusion, not an invented certainty.
Summary and recommendations
I would start with a material portfolio and a few well-supported impact mechanisms. Distinguish observations from estimates, connect scenarios to the timing of repayments and keep the route to a decision explainable. For each missing item, determine whether it prevents a conclusion, requires temporary caution or can be supplied later. The useful outcome is not a larger ESG database but a clearer understanding of a potential financial change and timely referral to the specialist responsible for evaluating it.