Which Sources Actually Ship Good Parts?

A lot of 3,000 microcontrollers clears incoming inspection on a Tuesday afternoon, or rather it doesn’t. The markings look right under the scope. The lab results come back the following week: blacktopping on a sample, and the decapsulation doesn’t match the manufacturer’s die. The lot goes back, if the source will take it. The customer’s order goes late either way.

Your inspector remembers that source. She flagged a lot from them last year for mismatched date codes. The buyer who placed this order didn’t know that. He saw a good price, a source that had shipped fine twice before, and a customer waiting.

The history existed. It was in a PDF inspection report from eleven months ago, in an email thread, and in your inspector’s head. None of it was on the screen when the purchase order went out.

I’ll be upfront about where I sit. I’ve never inspected a part or run a test lab. I spent close to 18 years running a digital agency, Infographic World, and now I use AI to fix real problems inside companies. What I’m good at is following a piece of information through a business and finding the spot where it stops moving. In a broker’s office, quality history is one of the clearest examples I know.

Where the quality history actually lives

Ask any broker which sources they trust and you’ll get a confident answer. Ask them to show the record behind it and things get vague. The facts are there. They’re spread across five places that don’t talk to each other.

• Inspection reports. Visual inspection, marking checks, X-ray, sometimes a full lab report. Usually a PDF per lot, saved to a folder named by PO number.
• Test results from outside labs. Another PDF, arriving days later, sometimes after the lot was already logged as received.
• Returns and customer complaints. In the ERP, as a credit memo or an RMA, often with a one-line reason.
• Shipping performance. Promised date against received date, in the purchasing module, if anyone filled in the promised date.
• Paperwork gaps. A missing certificate of conformance, a packing slip that doesn’t match the label, a date code different from the one quoted. These get noted in an email and then forgotten.

Nobody rolls this up because no one owns it. So buying runs on memory, and memory is generous to a source with a good price.

What the scorecard looks like

Read every record into one place

The first job is reading. The system goes through each inspection report, lab result, return and receiving note, pulls out the source, the part, the manufacturer, the date code, and the outcome, and files it against the source. A report that’s ambiguous gets flagged for your quality person to classify. It doesn’t guess whether a cosmetic note counts as a failure.

Build the scorecard by source and by part family

For each source, the scorecard shows lots bought, lots that passed inspection clean, lots with findings, lots that failed test, returns from customers, late shipments, and paperwork gaps. Each rate sits next to its count. It’s also broken down by part family, because a source that ships clean passives can still be the wrong place to buy programmable logic.

Your quality team sets what counts as a finding and how serious each kind is. The system applies their rules the same way every time.

Show it at the moment of purchase

A scorecard in a quarterly report changes nothing. One that appears as the buyer is about to issue a PO changes the decision. When a buyer selects a source for a part, the screen shows that source’s record for that part family, the last three lots, and any open issue. If the record crosses a line your quality team drew, the PO routes to them before it goes out.

The buyer still decides. Sometimes a source with a weak record is the only one with the part, and the right call is to buy with extra testing priced in. That judgment is theirs, made with the facts in front of them.

How this ties to AS6081

If you’re certified to AS6081, or working toward it, you already know how much weight it puts on how a distributor chooses its sources and on the records kept along the way. A rolled-up scorecard makes those records easier to produce and easier to defend. What the standard actually requires of your business, and what your auditor will accept, is for your quality manager and the standard’s own text to settle.

Run the numbers

The formula in words: the cost of one failed lot, times the number of failures a scorecard helps you avoid in a year.

The cost of a failed lot is more than the part. Add up test fees, return freight, the write-off when a source won’t refund, the credit or expedite cost to make the customer whole, and the hours your team spent on it.

Say, for the sake of the math, with numbers I made up:

• A failed lot costs $1,200 in test fees, $300 in freight, $1,500 in customer credit and expedite cost, and 15 staff hours at $60, which is $900. Call it $3,900 before any write-off.
• One in three failed lots ends in a write-off because the source won’t take it back, averaging $6,000. Spread across all failures, that adds $2,000 per lot.
• So the average failed lot costs about $5,900.
• You had 12 failed lots last year. Looking back, 4 came from sources that had already failed or shown findings on an earlier lot.

If a scorecard on the buyer’s screen had steered those 4 purchases elsewhere, that’s 4 × $5,900, or $23,600 a year. That leaves out the customer who quietly stops sending RFQs after a failed order.

To run it for real, pull three numbers from your own business:

1. Failed lots in the last 12 months. Count every lot that failed inspection or test, or came back from a customer.
2. Cost per failed lot. Pick three recent failures and add up every cost from the files, not from memory.
3. Repeat-source share. For each failed lot, check whether that source had a finding on an earlier lot. That share is the part a scorecard could have flagged.

Picture a Monday, before and after

This scene is made up.

Before. A buyer has a hot requirement for 800 pieces of a discontinued ADC. Three sources offer stock. He picks the middle price from a source he’s used twice. Nobody mentions that the inspection bench found remarked parts in that source’s last lot of a similar family. The lot fails test nine days later.

After. Same Monday, same three offers. When the buyer selects the middle-price source, the screen shows its record: five lots, two with findings, one failed test, all in data converters. The PO routes to the quality manager. She approves the third source instead, at a higher price, and adds a note to the scorecard explaining why.

What has to be in place first

• Inspection reports saved with the PO number and source name, so every report can be tied back to the purchase.
• A written definition of a finding from your quality team, with severity levels.
• Promised ship dates entered on every PO, or late shipments can’t be measured.
• Returns coded with a reason in the ERP.
• One owner for the scorecard, usually the quality manager, who reviews it monthly.

Guardrails worth keeping

• Evidence, not verdict. The scorecard informs the buyer and the quality team. It doesn’t block or approve a source on its own.
• Show the sample size. A rate without a count misleads.
• Keep it internal. A source’s record is your business’s private assessment. It isn’t something to share or publish.
• Let sources recover. Weight recent lots more than old ones, so a source that fixed its process can earn its way back.
• Log overrides. When a buyer goes ahead despite a weak record, the reason gets written down.

How you’ll know it’s working

• Failed lots per hundred lots received, quarter over quarter.
• Share of failures from sources that already had a finding. This should fall fastest.
• POs routed to quality review, and how many were changed.
• Cost of failures per quarter, using the full cost above.
• Time to pull a source’s history for an audit. It should drop from hours to minutes.

Where I’d start

Take last year’s failed lots and do the repeat-source check by hand. If a meaningful share came from sources that had already shown problems, you have your case. Then build the scorecard for one part family where failures hurt most, and put it on the buyer’s screen before you build anything else.

If you’d like to see this alongside a demand list built from your own requests, start with the free week on the Hunt page.