Your QC Reports Are a Supplier Scorecard: 6 Numbers to Track Every Month
Most importers file inspection reports in a folder and only open it when a container goes wrong. That is an expensive habit, because the reports you already pay for are the cheapest supplier performance dataset you will ever own — who is drifting, who fixes things fast, and who quietly needs a second pair of eyes.
Most importers file inspection reports in a folder and only open it when a container goes wrong. That is an expensive habit, because the reports you already pay for are the cheapest supplier performance dataset you will ever own — who is drifting, who fixes things fast, and who quietly needs a second pair of eyes. Order from China often enough and the same document you receive from Amazon FBA inspection China becomes a monthly scorecard, at no extra cost on top of the $169 man-day you already spend.
📊 Why a pass/fail line hides everything
“Pass” is a snapshot: one lot, one afternoon, one sample of 125 or 200 units against AQL 2.5. It answers a single question — did this batch clear the bar — and says nothing about direction of travel. A factory can pass six lots in a row while its average wall thickness walks from 0.85mm down to 0.72mm and its time-to-fix stretches from six days to nineteen. Both factories “pass”. Only one of them is failing on a timetable you cannot see yet.
We inspect across 2,000+ lots a year, and the pattern is boringly consistent: quality rarely falls off a cliff, it drifts. Our inspectors usually feel it first on the floor — “this carton is lighter than last month’s” — and reports come back typically within 24 hours, so the hunch can be turned into a number while the next purchase order is still open.
🧮 The six numbers worth tracking
| # | Number | How you record it | Why it moves money |
|---|---|---|---|
| 1 | Lot pass rate | PSI outcome per lot: pass / pass after rework / fail | Rework is where ship dates go to die |
| 2 | Majors per 100 units | majors found ÷ sample size × 100 | Compares a 125-unit sample with a 315-unit sample honestly |
| 3 | Criticals | Raw count, never averaged | One critical means stop and fix, not a defect rate |
| 4 | Time-to-fix | Days from report issue to rework signed off | Nineteen days in October eats a container booking |
| 5 | First-pass re-inspection yield | Share of reworked lots clean on the first re-check | A low yield means the fix was cosmetic |
| 6 | Drift on two key dimensions | Same measurement across the last three lots (thickness, GSM, slide travel, coating microns) | Catches silent material substitution before it reaches Amazon |
🚦 Where a number becomes a decision
One data point is noise; two in a row is a signal. These are the lines we go through when a client asks us to read their archive with them.
| Signal | Trigger | Action |
|---|---|---|
| Majors per 100 | Up in two consecutive lots | Add a during-production visit at the 30% mark of the next PO |
| Any critical | Count of 1 or more | Hold the release, switch that failure mode to a full check, ask for a written root cause in the next report |
| Time-to-fix | Over 14 days | Rework capacity problem: pull the next PO forward or split the order across two factories |
| Re-inspection yield | Under 80% | Treat the rework as cosmetic; require a corrective action note before releasing the balance payment |
| Key dimension drift | Two lots below the sealed-sample spec | Next lot gets measured against the golden sample, not against the drawing |
🛠 Three rules that keep the numbers honest
One lot, one row. A lot is one production run, one date code, one set of tooling. If the factory blended two runs into one container, that is two lots and two rows — otherwise the blended average hides the good half and the bad half.
One sampling plan. AQL 2.5 at inspection level II, same sample-size ladder every time. Change the plan and you are comparing noise with noise; the AQL calculator returns the sample size for your lot in one click so nobody has to guess on the factory floor.
Per 100 units, not raw counts. Raw defect counts reward whoever happened to get the larger sample. Normalising is what lets you put a 400-unit trial run next to a 6,000-unit production lot and still read the same story.
📁 The 20-minute monthly routine
Open the last quarter of reports. One row per lot, six columns, sorted by supplier. Look for two-in-a-row patterns rather than single bad afternoons, and take the sample size from the report rather than your memory. Keep the sheet next to your PO list — the value is in the comparison, not in the software. A spreadsheet is more than enough; CloudSpects clients run this off the PDF archive we deliver after every visit.
🎯 What the scorecard actually buys you
Price. A supplier at 0.4 majors per 100 with five-day fixes has a real argument for a higher unit price. The one at 3.2 majors and nineteen days does not — and now you can say so with a document instead of a feeling.
Priority. In the October-to-December crunch, factories allocate lines and QC staff to the buyers who bring data. “Your line 3 slipped twice last month” gets you a slot that a polite email does not.
Terms. Tie the last 20% of payment to a clean third-party report, then let the scorecard decide whether that report is routine paperwork or a warning light. Our inspection services and transparent pricing are built for repeat orders precisely because the second, third and tenth inspections tell you more than the first.
🇨🇳 Three suppliers, one quarter
A client buying 5,400 stainless steel kitchen racks a quarter across three factories had eleven reports sitting in a shared drive. We turned them into one sheet. Supplier A: 100% lot pass rate, 0.3 majors per 100, fixes signed off in four days. Supplier B: ten lots, majors per 100 walking 0.6 → 2.1 → 3.4, tube wall 0.80mm → 0.68mm, time-to-fix 11 → 18 days. Supplier C: four lots, one critical (missing fasteners in twelve cartons), then clean for three consecutive lots.
The decision was not to fire anyone. We moved 40% of B’s volume to A for the Q4 build, kept B on the non-load-bearing SKU where the thinner tube was acceptable, and added a during-production visit at the 30% mark. Across the following 5,400 units: 0.7 majors per 100, no criticals, longest rework three days. The tracking cost was part of inspection visits the buyer was already paying for, about $338 a quarter. The alternative — finding out from the review section in January — has no such price tag, which is exactly why importers keep paying it.
Not sure which two columns to start with? Send us your last three reports and we will read them with you before your next PO goes out.
❓ FAQs
What is a supplier scorecard in quality control?
It is a running record built from your inspection reports rather than from opinion: how often each factory passes, how many majors it produces per 100 units, how fast it fixes what it broke, and whether its key dimensions are drifting. Four to six lots is usually enough to see a trend worth acting on.
Which defects should count in a defect rate?
Count majors and criticals, and keep them in separate columns with the sample size from the report. Cosmetic minors are worth logging but they do not belong in the same number as a failed safety feature, or the average will hide the thing that actually stops a shipment.
Do I need software to track supplier performance?
No. A spreadsheet with one row per lot and six columns does the job, because the comparisons are simple and you look at them once a month. The failure mode is not a missing tool, it is reports that never get read twice.
How many lots before the numbers mean anything?
Around four to six lots per factory. Before that you are reading noise, so treat the first visits as a baseline and watch for two-in-a-row movement rather than a single bad afternoon on the factory floor.
Can I build this from reports I already have?
Yes, and it takes about twenty minutes. Pull the last quarter of reports, normalise everything per 100 units, and sort by supplier. Most importers find one factory whose numbers were quietly moving for months while every individual report said pass.
Frequently asked questions
What is a supplier scorecard in quality control?
It is a running record built from your inspection reports rather than from opinion: how often each factory passes, how many majors it produces per 100 units, how fast it fixes what it broke, and whether its key dimensions are drifting. Four to six lots is usually enough to see a trend worth acting on.
Which defects should count in a defect rate?
Count majors and criticals, and keep them in separate columns with the sample size from the report. Cosmetic minors are worth logging but they do not belong in the same number as a failed safety feature, or the average will hide the thing that actually stops a shipment.
Do I need software to track supplier performance?
No. A spreadsheet with one row per lot and six columns does the job, because the comparisons are simple and you look at them once a month. The failure mode is not a missing tool, it is reports that never get read twice.
How many lots before the numbers mean anything?
Around four to six lots per factory. Before that you are reading noise, so treat the first visits as a baseline and watch for two-in-a-row movement rather than a single bad afternoon on the factory floor.
Can I build this from reports I already have?
Yes, and it takes about twenty minutes. Pull the last quarter of reports, normalise everything per 100 units, and sort by supplier. Most importers find one factory whose numbers were quietly moving for months while every individual report said pass.