Build a supplier scorecard that separates delivery from quality

A supplier receives a strong performance score because most deliveries arrive on time. Quality control (QC) sees a different picture: repeated document discrepancies, several rejected receipts and an unresolved source-change notice. Both teams have accurate observations, but the combined score makes the quality concerns difficult to see.

A useful supplier scorecard preserves the differences between these questions. Did the supplier deliver as committed? Did the received material complete the required acceptance process? Were changes and investigations handled as agreed? The answers support different actions and should remain independently visible.

The purpose is to improve a supplier decision, not to produce a league table that looks precise while concealing its assumptions.

Choose the unit of analysis first

A supplier can serve several materials from different sites. Performance at the company level may hide a recurring problem in one source-material relationship. Start with the level at which the team can take meaningful action.

Keep the commercial supplier, manufacturing source, material and relevant receiving site available as dimensions. Then define whether each metric counts orders, order lines, shipments, receipts, lots or quantities. Those are not interchangeable denominators.

For example, one purchase order can arrive in three deliveries. Counting the order as on time because the first delivery arrived as promised gives a different result from measuring complete delivery of the required quantity. Neither definition should remain hidden in a report label.

For API businesses, supplier evaluation is addressed in ICH Q7 section 7. The specific scorecard design below is an original management method, not a prescribed regulatory scoring formula.

Keep requested and confirmed dates separate

The requested date tells the business when it wanted material. The supplier-confirmed date tells it what the supplier committed to. The actual receipt date records what happened.

A supplier can perform reliably against a date that is too late for the business. Conversely, procurement can request an unrealistic date and make an otherwise consistent supplier appear unreliable. Showing both comparisons helps distinguish the issues.

Preserve changes to confirmed dates. If the dashboard compares every delivery with the latest revised promise, a supplier can appear punctual despite repeated postponements. Record the original commitment, approved revisions and reasons so reviewers can see the effect of rescheduling.

Agree how partial quantities, early deliveries, non-working days and customer-caused delays are treated. Write those definitions next to the metric or in an accessible data dictionary. The report should be reproducible by someone outside the team that built it.

Work through the 40-receipt example

Consider an illustrative review period with 40 receipts. Of these, 36 arrived on time under the agreed definition. Thirty-two completed acceptance without an unresolved issue at the review cutoff.

The delivery rate is 36 divided by 40, or 90%. The second measure is 32 divided by 40, or 80%. It should be labelled according to its actual definition; it is not necessarily a rejection rate because some receipts may still be pending.

The overlap matters. Suppose 30 receipts were both on time and accepted without an unresolved issue. Then six were on time but had an unresolved issue, two were late but accepted without one, and two were late with an unresolved issue.

The combined favourable outcome is 30 divided by 40, or 75%, on this explicitly defined basis. Averaging 90% and 80% to obtain 85% would not establish that 85% of receipts satisfied both conditions.

This example does not prescribe a target. It demonstrates why the definition and intersection of measures matter before management interprets a score.

Distinguish pending work from adverse outcomes

A receipt awaiting testing is different from one rejected after investigation. A document discrepancy resolved before use is different from a confirmed material failure. Keep these categories separate so the review does not punish normal processing time or minimise a consequential event.

Use a stated cutoff date. Later quality decisions may change the outcome of a receipt included in an earlier review. Decide whether the organisation restates the earlier period, shows an updated view or retains both. Preserve the history so trends can be explained.

Do not convert missing data to a favourable result. A receipt without a linked disposition should appear as unresolved data until the responsible owner reconciles it. The count of missing links can itself reveal a process or integration problem.

Also separate severity and frequency. A frequent minor discrepancy and a rare consequential event require different attention. A weighted average can support a defined management process, but it should never make the underlying events disappear.

Add measures that lead to a decision

Choose a small number of measures tied to actual supplier-review actions. Useful candidates include delivery performance, quantity fulfilment, document completeness, accepted or rejected outcomes, open investigations and timeliness of required change communications.

For each measure, specify the owner, source records, calculation, exclusions and action it can trigger. If nobody can name a decision that would change, the metric may not deserve space on the main scorecard.

A recurring COA mismatch might lead to a document-process review. Repeated delivery postponements might lead to a planning or capacity discussion. An uncommunicated source change may require a quality assessment. These actions are more useful than telling the supplier to improve a composite score by five points.

The COA-verification guide explains document mismatch handling. The shortage-procurement guide shows how delivery uncertainty affects usable supply.

Reconcile data before ranking suppliers

Link purchase-order lines, receipt records and quality events using stable references. Check whether a single quality event affects several receipts or whether several events concern one lot. Avoid multiplying the same issue when joining tables.

Keep supplier identity mappings controlled. A renamed vendor or duplicate vendor code can split one supplier's history into apparently unrelated records. Conversely, grouping several manufacturing sites under a common commercial supplier can conceal the source of a problem.

Start with a manually reconciled sample. Compare the dashboard's counts with the underlying transactions and explain every difference. Automating an unreconciled query will make the report faster without making it more trustworthy.

Provide drill-through access appropriate to the reader. Procurement may need the event category and resolution status while detailed technical records remain within their approved access controls.

Run the review as an action meeting

Give reviewers the trends, material exceptions and previous actions before the meeting. Use the discussion to choose responses, owners and dates. Record why a measure changed and whether the earlier action worked.

Do not let the score automatically suspend or approve a supplier unless that behaviour has been explicitly defined and justified within the authorised procedure. The score is evidence for a decision; its interpretation depends on scope, severity, uncertainty and available alternatives.

AI can summarise repeated event themes or help classify inconsistent descriptions. Review the classifications and retain the source references. An unexplained risk label is less useful than a short list of traceable events that a supplier-quality owner can assess.

Summary and next actions

A supplier scorecard should make delivery, quality and responsiveness easier to understand independently and together. Its usefulness depends on definitions, reconciled records and decisions that follow the review.

• Choose one supplier-material relationship and define the receipt population and reporting period.
• Calculate delivery and quality measures manually, retaining their separate denominators.
• Separate pending work from adverse outcomes and inspect where delivery and quality issues overlap.
• Agree the supplier-review actions justified by the evidence, with an owner and follow-up point for each.

Connect the scorecard to supplier qualification. NextEdge Labs can help build the data joins, review views and action workflow while keeping the underlying evidence available to the people responsible for the decision.