A supplier scorecard is a recurring, data-based rating of each key supplier on the few dimensions that actually hurt when they slip: typically on-time delivery, quality, responsiveness, and cost. It exists to drive review conversations and sourcing decisions, not to decorate a slide.
Most plants have supplier scorecards in the sense that a spreadsheet exists. Far fewer have scorecards that change anything. The difference is not the template. It is a short list of metrics tied to real pain, a data feed people trust, and consequences attached to the score. This post covers the metrics worth using, honest definitions for each, how to weight and threshold them, and how to size the effort by supplier segment. A scorecard is one tool inside a broader supplier quality management program: the scorecard reports the relationship, the program manages it.
What metrics belong on a supplier scorecard?
Four to six metrics, each with an exact formula and a data source you trust. More than that and none of them get acted on. The usual core:
- OTIF (on-time in-full). The order arrived complete and on the agreed date; both conditions must hold. Nine lines of ten, delivered on time, is a miss. So is a complete shipment that lands three days late. Measure against the original promise date, not the most recent reschedule, or the metric flatters exactly the suppliers who slip.
- Defect PPM. Defective parts divided by parts received, times 1,000,000. If a supplier delivered 48,000 parts last quarter and incoming inspection plus line rejects caught 12 defectives, that is 12 ÷ 48,000 × 1,000,000 = 250 PPM. PPM makes small defect rates comparable across suppliers with very different volumes.
- Incoming inspection results and NCR count. How many non-conformance reports were written against the supplier, and how severe. A supplier can hold a decent PPM while generating a steady stream of paperwork, sorting, and returns that eats your quality team’s week.
- Responsiveness. Days to close a corrective action after you issue one (see CAPA), and quote turnaround time. A supplier who ships good parts but takes ninety days to answer a corrective action is telling you what the relationship will look like when something serious goes wrong.
- Price variance. Actual price paid versus quoted or standard price, tracked over time. It catches quiet creep: surcharges, minimum-order bumps, and expedite fees that never come up in the piece-price conversation.
| Metric | Formula | Where the data lives |
|---|---|---|
| OTIF | Orders delivered complete and on the original promise date ÷ total orders | ERP receipts vs. purchase order promise dates |
| Defect PPM | Defective parts ÷ parts received × 1,000,000 | Incoming inspection and line reject records |
| NCR count | Non-conformance reports issued against the supplier per period | Quality system NCR log |
| CA closure time | Days from corrective action issued to verified closure | CAPA log |
| Price variance | (Actual price − quoted price) ÷ quoted price | Purchase order and invoice history |
Put together, a scorecard for one supplier should fit on a page:
Supplier metrics belong next to your internal manufacturing KPIs for the same reason downtime belongs next to OEE: a late supplier and a slow changeover both show up as a line that is not running.
What does scorecard theater look like?
Twenty metrics nobody acts on, data assembled quarterly from systems nobody trusts, and scores that have never changed a single sourcing decision. Each failure has a tell. If the card carries a metric no one can define from memory, it is decoration. If receiving data is known to be wrong and people quietly work around it, the scores inherit the garbage. And if a supplier has been red for four straight quarters and still wins every purchase order it bids on, the scorecard is a ritual, and your suppliers can tell.
The fix is subtraction, not addition. Cut the card to the metrics someone will act on this quarter, and delete the rest without apology.
How should you weight metrics and set thresholds?
Weight by what hurts. For a critical machined component with one qualified source, quality takes the heaviest weight, because an escape costs a sort, a recall, or a line-down week. For commodity fasteners with five interchangeable sources, delivery and price can dominate, since a quality problem there is solved by switching. A common shape is quality 35%, delivery 35%, responsiveness 15%, cost 15%, adjusted by segment; the weights are yours to argue about, and the argument is the useful part.
Set red, yellow, and green thresholds per metric, in writing, before anyone sees a score: for example, OTIF green at 95% or better, yellow from 90 to 95, red below 90. Two cautions. First, thresholds are plant-specific; set them from your own history and tighten them over time rather than borrowing someone else’s benchmark. Second, act on the metric color, not the composite. A weighted composite is useful for trend, but it can average a red quality quarter into a green overall score, and quality reds do not average away on the receiving dock.
How often should you review supplier scorecards?
Collect the data monthly, hold quarterly business reviews with critical suppliers, and review the rest annually. Cadence should follow segmentation, and segmentation comes down to two questions: how much do we spend with this supplier, and how badly does it hurt if supply stops?
Effort follows the quadrant. Strategic suppliers get monthly data and a quarterly review with named actions on both sides. Bottleneck suppliers, low spend but painful to lose, get watched for risk, and the real action is qualifying a second source. Leverage suppliers get competition. Transactional suppliers get automation and an annual glance, and that is enough.
Unreliable supply also has a visible carrying cost. U.S. manufacturers were holding inventories at about 1.5 times monthly shipments through 2026, with an inventories-to-shipments ratio of 1.47 in May 2026 per the Census Bureau’s M3 survey: roughly six weeks of shipments sitting as inventory. Some of that is deliberate strategy, but a share of it is buffer against suppliers who cannot be trusted to hit a date. Every point of OTIF a supplier recovers is safety stock you no longer have to carry.
How do you build a supplier scorecard?
- Pick six or fewer metrics tied to real pain. If the line stopped twice this year on supplier shortages, OTIF is on the card. If incoming rejects are eating inspection hours, PPM and NCR count are on. If nothing hurts, that supplier may not need a scorecard at all.
- Define every formula and data source in writing. Decide now how partial shipments, customer-caused reschedules, and drop-shipments count. Undefined edge cases become arguments in the review meeting.
- Set weights and red/yellow/green thresholds by segment. Quality heaviest for critical parts; delivery and price heavier for commodities. Publish the thresholds before the first scores go out.
- Automate the data pull. If a buyer spends two days a quarter assembling the scorecard by hand, it will be quietly abandoned within a year. The card should build itself from receipts, inspection records, and the NCR log.
- Set the review cadence by segment. Monthly data for everyone. Quarterly business reviews for strategic and bottleneck suppliers, with the scorecard as the agenda. Annual for the rest.
- Tie scores to consequences. Green earns business share. Yellow gets a conversation. Red gets a development plan with dates, shifted volume, or an exit path. A scorecard without consequences is a newsletter.
The step that kills most scorecards is the fourth one: the data lives in the ERP, the quality log, inspection spreadsheets, and email, and nobody owns the stitching. Harmony is an AI-native layer that connects ERP, QMS, spreadsheets, and paperwork into one operational layer, with no rip-and-replace, so receipts, NCRs, and corrective-action records can feed one live view instead of a quarterly copy-paste. AI search answers questions like “every NCR against this supplier in the last year” with cited sources, and workflow automation can fire a notification or a QMS log entry the moment a metric crosses a threshold, instead of waiting for the review meeting to notice.