How to Build a Supplier Scorecard

The common failure is not a bad metric. It is a scorecard nobody acts on.

A score gets calculated monthly, filed, and referenced only when a relationship is already in trouble. The supplier either never sees it or sees it once a year in a slide with no path to changing it. Nothing about the way orders are placed changes as a result. At that point the scorecard is bookkeeping.

The second failure is measuring too much. Fifteen weighted metrics feel rigorous and produce a single number that moves for reasons nobody can explain. If a metric cannot change a decision you would make, it is noise that dilutes the metrics that can.

A scorecard earns its keep when it does one of three things: shifts volume between suppliers, triggers a corrective action with a due date, or justifies developing a second source. If your draft does none of those, cut it down until it does.

One thing to be clear about before you start. A scorecard is not a spreadsheet, it is a program: data captured at receiving on every order, a definition both sides signed up to, a review on the calendar, and someone who owns the follow up. Each of those is small. Together they are a standing commitment, and the last section of this article puts a number on it.

Quality, delivery, responsiveness and cost. In most supply bases these four explain nearly everything you care about, and adding more mostly adds argument.

Quality is the outcome you cannot buy back. Measure it as parts per million defective if your volumes support it, or as lot acceptance rate if they do not. For low volume work, lot acceptance is the honest measure: five rejected lots out of forty tells you more than a PPM figure computed on 300 pieces.

Delivery is the one people think is simple. It is not, and the next section is about why.

Responsiveness is underrated and easy to capture. Time from RFQ to quote, and time from issue raised to a substantive answer. It predicts how a supplier will behave in the week something goes wrong, which is exactly when you need to know.

Cost performance measures movement against a baseline rather than absolute price. A supplier that is not the cheapest but reduces cost year on year through real changes is worth more than one that is cheapest today and static.

On time delivery is where scorecards go wrong most often, because both sides believe they are measuring the same thing and they are not.

Agree four things in writing. Which date counts: the date the supplier acknowledged, or the date you requested. Almost every dispute traces back to this one. What the window is: same day only, or a tolerance such as two days early to zero days late. What event marks delivery: shipment or receipt. And what a partial shipment counts as, since a line received at eighty percent is not obviously on time or late.

Write the definition into the purchase order terms. A supplier being measured against a rule it has not seen will not treat the score as legitimate, and it will be right not to.

One more thing worth doing: track the reason for lateness even if the score does not weight it. A supplier late because you released a drawing revision mid job is a different problem from one late because it took the work without capacity, and a score that cannot tell them apart will eventually punish the wrong supplier.

Convert each measure to a common scale before weighting. Bands are easier to defend than a continuous formula, and easier for a supplier to act on.

Set the thresholds where your business feels the difference. If ninety two percent on time delivery is materially worse for you than ninety six, do not use bands ten points wide. If it is not, do not pretend it is.

The number to watch is not the composite, it is the trend. A supplier at 78 and rising is in better shape than one at 88 and falling, and a scorecard that only reports this month makes both look like a single number.

Quarterly is the right default for most supply bases. Monthly for a supplier in corrective action or in a ramp. Annually for suppliers whose spend does not justify the effort, which will be most of them.

Send the score before the meeting, not during it. A supplier reading a number for the first time in a review will spend the meeting defending rather than fixing, and you will learn nothing.

Have the person who can change something in the room. A scorecard reviewed with a sales contact who cannot move a schedule or open a machine is theater. On the supplier side that usually means operations or quality; on your side it means whoever places the orders, because the scorecard has to be connected to that decision to have teeth.

Close every conditional or at risk score with a written action, an owner and a date. An action item without a date is a note.

That is a real block of calendar. A quarterly review done properly is thirty minutes of preparation, thirty in the room and an hour of follow up, per supplier, four times a year, plus whatever it took to capture the data in the first place.

If you are a small hardware team, most of the above is more machinery than your situation supports, and the honest reduction is a spreadsheet with four columns.

Per purchase order, record: was the lot accepted, was it received inside the agreed window, how long did the quote take, and did the paperwork you asked for arrive with the parts. That is four fields per order, filled in at receiving, and after ten orders per supplier you have a real pattern rather than an impression.

Skip weighting until you have enough orders for a composite to mean anything. Read the columns directly. A supplier with a perfect quality column and a poor delivery column needs a scheduling conversation, and averaging those two into a 74 hides exactly the thing you would have acted on.

The value at this size is not the score. It is that you notice the pattern in month three rather than month nine, and that when you do raise it you are describing eleven orders rather than a feeling.

Decide this in advance, because a scorecard with no consequence attached teaches everyone to ignore it.

The usual set: eligibility for new programs, share of volume where a part is dual sourced, whether a supplier is invited to quote new work, and whether a corrective action escalates to a development plan or a transition. Some organizations tie payment terms or a price review to it, which works when the spend is large enough to matter to the supplier.

What a score should not do is trigger an automatic exit. Sourcing a replacement for a machined part with a qualified process, tooling and a FAIR behind it costs more than most scorecards account for, and a supplier at 68 who is fixable is usually cheaper than a new supplier at an unknown number.

The scorecard tells you where to spend attention. The decision about whether to spend it stays yours.

Nobody budgets for this, which is why scorecard programs are started often and sustained rarely. The work is not hard, it is just continuous, and it lands on whoever is already the busiest person in a small hardware team.

Four things consume the time. Capturing data at receiving on every order. Preparing the score and reconciling the inevitable disagreement about which dates counted. The review itself. And the follow up, which is where the value is and which is the first thing dropped when a build slips.

The figures below assume you already inspect incoming parts and already record receipts, so they count only the scorecard work on top. They do not include sourcing and qualifying a replacement when a supplier fails, which is the expensive event the program exists to see coming.

Everything above assumes you are going to hold supplier relationships directly. That is one of two structures, and it is worth choosing deliberately rather than ending up in it.

The other is to source through a partner that holds the supply base for you. You send the part; the shops, the qualification, the quality documentation and the follow up when something is late sit on the other side of one relationship. Your scorecard becomes a scorecard of one, which takes about twenty minutes a quarter and does not decay when a build slips.

This is what OpenSpindle does. We hold a network of independent machine shops across processes and regions, place your part with one that can run it, and carry the certifications, material certificates and inspection reports so you can ask for them rather than chase them. Quotes come back in about 48 hours. When a part changes, or a process you have never bought before turns up on the BOM, you do not start a qualification cycle, you send the file.

Be clear about what you give up, because it is real. You are not choosing the shop by name, you are specifying the requirement and letting it be matched. For a part whose supplier relationship is itself strategic, or where volume is high enough that a few points of unit price outweighs everything else, holding that relationship directly is the right answer and the program above is how you do it well.