← Insights

Suppliers

Supplier scorecards that run on data, not anecdotes

July 2026 · 7 min read

Sit in enough supplier review meetings and you notice the same physics: the supplier who caused last week's line stop dominates the hour, the supplier who quietly delivered perfectly for six months is never mentioned, and decisions get made on whichever anecdote arrived most recently and loudest. Recency and volume are not data. A scorecard exists to replace them with something a purchasing decision can safely stand on.

Having spent years auditing and managing suppliers across continents, I can tell you the difference between panels managed by anecdote and panels managed by scorecard is visible within two quarters. Here is how to build one that works.

Five criteria cover almost everything

Resist the twelve-criterion monster; it measures nothing well. Five dimensions capture what matters for most manufacturing supply bases:

• Quality - defect rates, NCRs raised, returns. The heavyweight.
• Delivery - on-time-in-full against the promise, not against the renegotiated promise.
• Cost - competitiveness and price stability, including the invoicing noise that consumes your accounts team.
• Responsiveness - speed and quality of communication when something goes wrong: 8D responses, corrective action turnaround, engineering queries.
• Compliance - certificates current, documentation right first time, audit findings closed on time.

Score each 1-5 monthly, against a written guide: what does a 5 for delivery mean (say, 98%+ OTIF), what does a 2 mean? The written anchors are what let two different reviewers produce the same score - the same trick that makes a 5S audit comparable month to month.

Weight them - because the criteria are not equal

A naive average says quality matters exactly as much as invoicing tidiness. No factory actually believes that. Weight the criteria - a typical manufacturing profile runs quality 30%, delivery 25%, cost 20%, responsiveness 15%, compliance 10% - and make the weights an explicit, recorded decision that sums to 100. The weighted result on a 0-100 scale turns into a grade: A (85+) preferred, B (70-84) acceptable with watch items, C (below 70) formal improvement plan or exit path.

The grade is not the punishment; it is the trigger. An A earns more business. A C earns a documented improvement plan with dates. What a C must never earn is a shrug and another identical month.

The trend is worth more than the number

A single month's score tells you almost nothing - one bad shipment can dent a good supplier's month. Twelve months in a row tell you nearly everything:

• A supplier sitting at 88-92 all year is a partner. Send them more work.
• A supplier climbing from 61 to 78 over six months is an improvement plan actually working - which is evidence about your supplier development process, not just about them.
• A supplier sliding 85, 82, 78, 74 is the next line stop announcing itself politely, months in advance. This slide is invisible in anecdote-world, because no single month was bad enough to become a story.

Lay the year out as a heat row per supplier - red through green - and the trend reads at a glance: the improver, the slider and the steady performer identify themselves without anyone constructing a narrative.

Make it survive contact with reality

Feed it from records, not memory. Quality scores come from the NCR log, delivery from goods-in receipts. Scoring from memory in the review meeting reintroduces exactly the recency bias the scorecard exists to remove.

Score the panel, not just the problems. The quiet excellent supplier deserves to be discovered - they are your benchmark and possibly your dual-source candidate.

Share the scores with the suppliers. A scorecard the supplier never sees is surveillance. Shared quarterly, it becomes the agenda of the review meeting and the fairest possible basis for the difficult conversations - and ISO 9001 clause 8.4 asks for exactly this kind of defined evaluation, monitoring and re-evaluation criteria anyway.