ESRS Estimates and ‘Undue Cost or Effort’: When Relief Is Defensible

Technical status note. This article is based on C(2026) 5010 final, adopted 3 July 2026. The final Official Journal version and any implementation guidance issued after 2 August 2026 remain update triggers.

Use note. “Undue cost or effort” is a contextual reporting judgement, not a generic exemption. The examples and registers below show a defensible process; they do not predetermine the outcome for a specific undertaking, metric or reporting period.

Data gaps are normal in sustainability reporting. They arise because the information is outside the financial close, located in operational systems, held by suppliers, estimated through scientific models or not yet governed with finance-level controls. ESRS accommodates this reality through estimates, measurement-uncertainty disclosures, partial-scope reliefs and the undue-cost-or-effort principle.

The risk is turning flexibility into an undocumented absence. A spreadsheet cell marked “N/A — undue effort” does not show what information was sought, which alternatives were tested, why the result would not be useful or how the gap will improve. Assurance teams will usually ask for the decision trail, not merely the conclusion.

A strong approach treats estimates as controlled information: defined method, source hierarchy, assumptions, calculation, sensitivity, ownership, review and annual reassessment.

Reasonable estimates can be useful information

ESRS 1 AR 42 states that reasonable estimates, including scenario or sensitivity analysis, are an essential part of preparing the sustainability statement. High measurement uncertainty does not automatically make information useless if significant assumptions and estimates are explained.

This is a critical distinction. An estimate is not weaker merely because it is an estimate. Its usefulness depends on whether it is relevant, faithfully represented, transparent about uncertainty and based on a method proportionate to the decision.

Significant judgements and uncertainty must be visible

Paragraph 88 requires information enabling users to understand the judgements with the most significant effect on reported information, significant uncertainties including reliance on estimates, and significant assumptions and limitations. The most difficult, subjective or complex judgements deserve the clearest evidence trail.

The information search is proportionate, not exhaustive

Paragraphs 93–95 require all reasonable and supportable information available without undue cost or effort. AR 45 explains that this includes undertaking-specific and external information about past events, current conditions and future forecasts, but does not require an exhaustive search.

Information expected to be considered includes:

risk management processes;

information used to prepare financial statements;

information used to operate the business model and set strategy;

sustainability due-diligence records;

information used to manage material IROs;

sector or peer-group experience;

scientific research and other available external sources.

The assessment is specific to the undertaking

Paragraph 94 requires a balanced consideration of cost to the undertaking and benefit to users. Relevant factors include size, resources, technical readiness, value-chain scale and complexity, and the availability of digital information-sharing tools. A method that would be proportionate for a global bank may be excessive for a first-time mid-sized reporter; the reverse can also be true where a small entity operates a high-impact site with readily available direct data.

Availability should improve over time

Paragraph 95 requires reassessment each reporting period and expects information availability to reflect prior improvement actions and greater external data availability. The same unsupported “undue effort” conclusion repeated for several years is difficult to defend unless circumstances genuinely remain unchanged and the undertaking can explain why improvement is not feasible.

Figure 1. A defensible undue-cost-or-effort conclusion is the final stage of an evidence funnel.

Stage 1. Define the information need

Start with the material IRO, disclosure objective and required quality of information. Ask:

What decision should the information support?

Is the output a materiality conclusion, a narrative, a metric or a financial effect?

What boundary, period, unit and level of aggregation are necessary?

Does the applicable ESRS provide a specific relief or prohibit partial scope?

Without a defined information need, the team cannot assess whether an alternative is useful. A request for “exact supplier emissions” is too broad; “an estimate of Category 1 emissions sufficiently reliable to identify significant categories and explain uncertainty” is testable.

Stage 2. Inventory reasonable and supportable information

Create a source log before sending new requests. Classify sources as:

direct internal operational data;

financial and transaction data;

direct counterparty or site data;

internal model inputs and prior-year information;

external scientific, regulatory, sector or peer information;

proxy data, factors and samples.

Record availability date, owner, coverage, quality, restrictions and whether the information is already used in a business or finance decision. Information already used to price, budget, manage risk or meet a legal requirement will often be difficult to dismiss as unavailable without undue cost or effort.

Stage 3. Test alternative methods

Before concluding that further information would involve undue cost or effort, assess feasible alternatives. Depending on the requirement, these may include:

estimation from activity or spend data;

representative samples;

supplier-category or geographic proxies;

ranges instead of single points;

non-monetary quantities;

combined financial effects;

a clearly defined partial reporting scope where permitted;

aggregation that preserves decision-useful information;

sensitivity analysis or scenario bands;

an entity-specific metric where the topical standard does not provide one.

Each rejected alternative should have a reason: insufficient relevance, unreliable relationship to the underlying phenomenon, double counting, unacceptable bias, lack of a defined population or cost disproportionate to user benefit.

Stage 4. Make the balanced cost-benefit judgement

The judgement register should compare the incremental cost and effort of the next-best method with the expected information benefit. Costs can include supplier burden, specialist modelling, system changes, verification, data licensing and management time. Benefits include the ability to understand a material IRO, compare periods, connect to financial effects, monitor a target or avoid misleading boundary omissions.

The assessment should not be reduced to a monetary threshold. Consider severity of impacts, potential magnitude of financial effects, uncertainty, decision sensitivity, user need and whether the information is central to a public target or claim.

Stage 5. Report a transparent output

The conclusion may be direct data, an estimate, a range, a combined effect, a permitted partial scope or a qualitative disclosure under a quantitative relief. In every case, explain the method and material limitations. Where a relief requires it, disclose why quantitative information was not provided, identify affected financial-statement line items and provide combined quantitative information unless that too would not be useful.

The accompanying Excel workbook includes a JUDGEMENT REGISTER. A robust record contains:

Figure 2. Estimates should move through a repeated define–source–estimate–control–disclose–improve cycle.

A practical hierarchy avoids choosing the easiest factor first:

Direct measured or recorded data. Meter readings, payroll, invoices, production data, incident records and recognised financial amounts.

Direct calculated data. Activity data multiplied by controlled factors, supplier-specific calculations or engineering models.

Representative samples. Controlled selection with a defined population, coverage rationale and extrapolation method.

Entity- or category-specific proxies. Similar sites, products, suppliers or customer groups with documented comparability.

External averages and generic factors. Sector, geography, spend or lifecycle factors, with limitations and sensitivity.

Qualitative or range-based information. Used where a single numerical estimate would not be decision-useful and the relevant ESRS permits it.

Minimum controls should include version-controlled methodology, source retention, unit and boundary checks, reconciliation to finance or operational totals, independent review, change control, sensitivity analysis for material assumptions and approval of material estimate changes.

Context. A food manufacturer identifies a material water-related impact in two stressed basins. It purchases agricultural inputs from 2,400 farms. Direct withdrawal and consumption data are available for 180 large farms covering 48% of purchases; the rest have no meters or common reporting system.

Information need. The undertaking needs decision-useful information about water consumption connected with the material supply categories and basins, not an exact farm-by-farm number for every supplier.

Sources considered. Purchase volumes, crop type, basin, irrigated area, direct data from large farms, government crop-water factors, satellite-derived irrigation information and scientific research.

Alternatives. The team tests a purchase-volume proxy, crop-and-basin factors, a stratified sample and a partial scope. A pure spend proxy is rejected because price variation is poorly related to water use. A 100% direct survey is judged impracticable and unlikely to produce reliable data in the reporting timetable. A stratified estimate using direct data for large farms and crop-basin factors for the remainder is selected.

Controls. The model reconciles purchases to procurement, separates irrigated and rain-fed crops, applies factor version control and includes sensitivity ranges. A hydrology specialist reviews the method. The disclosure explains 48% direct coverage, the estimated portion, assumptions and uncertainty.

Improvement plan. Direct data collection expands to high-volume farms in the two most stressed basins, targeting 70% purchase coverage next year. The undertaking will reassess whether the selected factor set remains available without undue cost or effort.

Why it works. It identifies the information need, direct coverage, estimate method, rejected alternative, uncertainty, review and improvement plan. It does not suggest that the estimate is exact or that the data gap justifies excluding the supply category.

What disclosure objective and material IRO does the estimate support?

Which internal information already used by management or finance was considered?

Which external sources were available at the reporting date?

Why was the search reasonable but not exhaustive?

Which alternative methods were tested and why were they rejected?

How was cost and effort balanced against user benefit?

Does a specific ESRS relief apply, and are all its conditions met?

Is the reporting scope objectively defined and reconciled to the full population?

What are the significant assumptions, uncertainties and sensitivity drivers?

How were the method, calculation and disclosures reviewed?

What improvement actions were promised previously, and were they delivered?

Could the same conclusion be reached consistently by an independent reviewer from the retained file?

Is an estimate allowed when direct data could theoretically be collected?

Potentially. ESRS permits estimates depending on practicability, reliability and the reasonable-and-supportable-information principle. The undertaking should explain why the selected estimate is decision-useful and why further direct collection would involve undue cost or effort relative to benefit.

Must the undertaking contact every value-chain actor?

No. ESRS does not require an exhaustive search or information on every actor. The collection and estimation strategy should be driven by material IROs, representative coverage and reliable methods.

Can a metric be reported for only part of own operations?

For certain metrics, paragraph 91 permits an objectively defined partial scope where reliable direct or estimated data are available only for that part, with disclosure and an improvement plan. The relief does not apply to E1-8 gross Scope 1, 2 and 3 emissions.

Does high uncertainty always justify qualitative disclosure only?

No. High uncertainty does not automatically make quantitative information useless. The applicable relief test must be met, and ranges, sensitivity or combined effects may still provide useful information.

How often should the judgement be revisited?

At least each reporting period, and sooner if a new system, dataset, methodology, acquisition, material IRO or external factor changes the information available or the cost-benefit balance.