Supplier Ethics Metrics: Seven Requests for Any Report

Almost every number in supply chain reporting comes from someone with an interest in it. A reader cannot fix that by reading harder. The fix is to ask for different numbers.

Seven requests, made before the report is accepted, turn a table of counts into something a buyer can use. Send them in writing to whoever produces the report, whether a vendor or your own team.

What to take away

Request one: announced against unannounced, same sites

Ask for both kinds of visit at every site that has had both, side by side. If the two are similar, your announced program is seeing roughly what is there, which is not the usual result.

If they diverge, the size of the divergence measures what preparation removes, and it applies to every site you have only visited by appointment. This is the single most informative number available to a buyer, and it requires only that you visit a sample without warning.

In the US, California's Transparency in Supply Chains Act requires a covered company's disclosure to say if its supplier audits were not independent, unannounced audits. Disclosure is not proof of conditions; the useful question remains what the company has established, rather than what it has collected.

Request two: recurrence at the next visit

Ask what share of last visit's findings were present again. Closed corrective actions are a throughput measure and describe paperwork. Recurrence shows whether conditions moved.

A high recurrence rate with a high closure rate is the signature of a program that is processing paperwork, not removing the problem.

Request three: received volume against plausible capacity

Ask for declared headcount, shift pattern and equipment set against what you actually took delivery of. If deliveries grow while declared headcount, shifts and equipment do not, investigate what changed before treating audit completion as reassurance.

This check cannot be prepared for, because it uses your own records. When it does not reconcile, the difference was made somewhere you have not assessed, and no certificate covers it.

Request four: a provenance label on every figure

Ask for a column that says who observed or declared each number. Provenance decides how much a figure can carry, and most reports put all four rows below in one table with no label for which is which.

The label is a one-line change that changes how the whole document is read. Declared figures are useful mainly where they fail to reconcile with the observed ones.

Request five: what each count really shows

Ask the author to replace each heading with what the line actually establishes. The claim then shrinks to its true size.

Supplier reports may cite programs, standards, platforms or certifications such as SMETA, amfori BSCI, SA8000, Sedex and Fairtrade. The name alone does not establish current conditions across the chain. Ask what was assessed, at which suppliers, and when.

Request six: the denominator under every rate

Ask what the base was for every figure. Counts move with the size of the base, so a program that adds suppliers will show rising findings and rising grievances without anything changing in the world.

Prefer rates with a stated denominator: findings per site visited, grievances per thousand workers, recurrence as a proportion of prior findings. The denominator belongs in the report itself, because a rate whose base is unexplained is less honest than a count.

Request seven: the distribution, and what changed in the method

Ask for the spread across suppliers, and the worst one by name. Supply chain problems are local and concentrated. A chain-level average with a stable trend routinely conceals one supplier where something is badly wrong, so look for the distribution and the outlier before the mean.

Then ask whether the definition, the population or the visibility of reporting changed since last time, and consider underlying conditions only after those. Visibility changes move these numbers hardest and are mentioned least often. A program that has just built a worker channel and reads the resulting rise as deterioration will suppress the mechanism it paid for.

What a good answer looks like

The Justice Department's federal prosecution principles for business organizations tell prosecutors to determine whether a compliance program is merely a paper program or was designed, implemented and resourced to work in an effective manner. The same test applies to a metrics report.

GRI Standards, IFRS Sustainability Disclosure Standards, SASB Standards, and the UN Guiding Principles Reporting Framework structure reporting. They do not independently verify supplier-level claims.

What none of this measures

Whether the people making your products are better off than they were. No available metric answers that, and programs that claim to are usually reporting their own activity.

What the measures above can establish is narrower and still worth having: whether you can see anything, whether last time's findings got fixed, and whether affected people have a route to tell you.

Common questions

Our suppliers give us hours and wage data. Is it worthless?

Not worthless, but it is a declaration. Its main value is in reconciliation: when payroll, output and time records do not agree with each other, that inconsistency is a real finding even though each document on its own is unverified.

How many suppliers should we measure?

The ones in your highest-risk category, which is usually a small number. A thin measure across every supplier produces a table nobody acts on; three comparisons on ten suppliers produce decisions.

Should we publish these numbers?

Publish what you can evidence, with provenance marked and the things that got worse included. A report where every line improved is describing a selection, and readers who know the subject can tell.