Supplier reporting coverage measures and their limitations
A supplier reporting percentage can answer a narrow question precisely while leaving a broader question unresolved. This study asks when a disclosed measure identifies coverage of a selected group, and what a reader can legitimately infer about everyone outside that group. The distinction matters before comparing percentages, interpreting a complete response claim or deciding which additional evidence would reduce uncertainty.
We examine three disclosure cases in three UK public-sector publications: two UK government statements and one Nuclear Waste Services statement. These are purposively selected historical observations, not three independent company samples or a time series. Each provides a different counting or selection problem.
Case one: a defined invitation population
The UK government’s 2019–2020 statement reports that 295 suppliers on 26 high- or medium-risk commercial frameworks were invited to the Modern Slavery Assessment Tool in autumn 2019; 210 had completed it at publication. Our calculation is 210 ÷ 295 × 100 = 71.2%, rounded to one decimal. That describes completion among those invited. The stated £2.8 billion annual spend belongs to the invited group; the passage does not allocate spend to completers. It therefore cannot establish a spend-weighted completion percentage.
Case two: eligibility inside a selected group
Nuclear Waste Services’ 2025–2026 statement includes a 2024–2025 table row: 28 suppliers represent 80% of spend, 19 qualify to publish a modern slavery statement, and 100% of the qualifying suppliers have published. The percentage is conditional on the qualifying group within the selected spend population. It does not say that all 28 published, or that all suppliers published. We use this historical row only; we do not combine its counts with another year’s supply-chain total.
Case three: adjacent counts with different coverage
The UK government’s 2021–2022 statement reports 360 high- or medium-risk suppliers identified by central departments, then 939 supplier responses across government including partner organisations. It reports that 84% accepted recommendations and completed their campaigns. These descriptions do not establish one shared invitation denominator. Dividing 939 by 360 would create a misleading response rate. Nor does the passage establish that every response represents a different supplier. The rounded percentage cannot be reversed into an exact completion count.
What the cases establish
The cases distinguish invitation completion, publication within an eligible subset, and campaign progress across a differently described population. They support separate readings, not a league table of supplier engagement.
Our method records the counted object, selection condition, reporting date or period, numerator, denominator and permitted inference. We selected accessible official passages containing quantitative supplier-reporting information and enough context to examine at least one of those relationships. We excluded general supplier commitments, unsupported estimates of the full supply base and environmental quantities that did not measure reporting coverage. We did not check individual suppliers’ submissions or assess the underlying programmes.
Read the noun after the number. A supplier, a response, a campaign and a published statement can be different counting objects. Then read the condition: invited, eligible, selected by spend, or included through partner organisations. A mathematically valid division still needs compatible populations and units. If those conditions are missing, preserve the unresolved question rather than manufacture a percentage from nearby figures.
Original experiment: one complete subset, many possible totals
The following experiment is entirely fictional. A buyer has a fixed roster of 20 suppliers, A01 to A20, at one review date. Eight are selected for a document-publication check. Of those eight, five meet the exercise’s stated eligibility rule. All five have the specified document available. The publication status of the other 15 suppliers is unknown. Eligibility here is an invented study rule, not a statement of anyone’s legal duty.
For the eligible selected group, the supported result is 5 ÷ 5 = 100%. Changing the question to “What share of all 20 suppliers has the specified document available?” requires information the check did not collect. The five known publications give a lower bound of 5 ÷ 20 = 25%. If every unknown supplier also has the document, the upper bound is 20 ÷ 20 = 100%. Neither endpoint is an estimate of the actual total.
Two possible worlds make the limitation visible. In World L, only the five observed suppliers have the document. In World H, every supplier has it. Both worlds produce the same 100% result for the eligible selected group. Because the observed result fits both worlds, it cannot distinguish them. This is an identification test: it asks what the stated evidence rules out, without assigning probabilities to the unobserved cases.
The experiment can be reproduced by assigning either available or unavailable to each of the 15 unknown suppliers. There are 2 to the power of 15, or 32,768, possible assignments. Every assignment leaves the observed five unchanged. The resulting whole-roster counts run from five to 20. Treating all assignments as equally likely would introduce an extra assumption; our enumeration makes no such probability claim.
Choose evidence that narrows the unanswered question
Suppose a further check confirms that A09 and A10 have the document and that A11 does not, under the same document definition and review date. There are now seven confirmed publications, one confirmed absence and 12 unknowns. Whole-roster coverage is bounded by 7 ÷ 20 = 35% and 19 ÷ 20 = 95%. The range narrows because additional states are known. The original five-of-five measure remains 100%.
The confirmed absence is a stipulated observation in this example, not merely an unsuccessful website search. In real research, failure to locate a document may leave its existence unresolved. Record what the check actually established. Likewise, a document found after the review date might clarify earlier availability or represent a later publication; those are different changes and should not be silently combined.
The change from a 25% lower bound to 35% does not demonstrate that any supplier improved. In the experiment, all statuses relate to the same date; the researcher learned more. Distinguishing new knowledge from changed behaviour prevents a reporting update from becoming an unsupported performance claim. A narrower range can be useful even when the actual coverage remains unidentified.
The next research step follows the unresolved question. If the reader needs whole-roster publication coverage, check the remaining unknowns against the same definition. If the reader only needs completion within the original eligible subset, that calculation was already identified. Neither task automatically requires obtaining confidential supplier records or widening the programme’s purpose. Evidence collection should serve the stated question.
Keep coverage separate from the outcome it cannot measure
Publication coverage records whether a defined document is available. It does not establish whether every statement is accurate, whether the underlying practices are effective, or whether every tier of a supply chain has been examined. A complete selected group may be a useful administrative result while those substantive questions remain open. The study makes no finding of misconduct, compliance or assurance about any supplier or reporting body.
For a research note, preserve the shortest sentence that keeps the selection condition attached to the result. State the population, the counted object and the reference date; distinguish a calculated percentage from one reported by the source. If the wider total is unknown, name it explicitly. Do not replace an eligibility-conditioned result with a general claim that supplier reporting is complete.
The practical collection workflow is a separate task. Nitium’s supplier response tracker records individual request exchanges, evidence and coverage reviews. This study instead examines what aggregate public descriptions identify and demonstrates the possible totals that remain consistent with incomplete observations. It offers no benchmark and does not represent Indian suppliers, all UK public procurement or supplier sustainability reporting generally.
Sources checked 15 October 2026. Case one: UK government 2019–2020 statement, common-goods-and-services supplier subsection.
Case two: Nuclear Waste Services 2025–2026 statement, Figure 2, 2024–2025 row only.
Case three: UK government 2021–2022 statement, Section III, risk assessment and due diligence.
Contains public-sector information licensed under the Open Government Licence v3.0. Analysis, fictional data and artwork are original Nitium material. No source photographs, logos or graphics are reproduced, and no endorsement is implied.
Comments