A Supplier Scorecard for Display Manufacturers: Weighting Quality, Response and Continuity

The shortlist is down to four suppliers, the samples are on the bench, and someone has to write the recommendation. That is the moment most teams open a spreadsheet, invent six columns, score from one to five and total the rows. The arithmetic is easy, which is why nobody questions it. What the total hides is that the columns are not equally important, that their importance is not constant, and that one decisive event can make a high total worthless.

A workable LCD supplier scorecard has three properties. The weights follow the product life stage, responsiveness is measured in days rather than adjectives, and price is normalised to delivered cost before it is scored. The sections below set out the criteria worth scoring, the weights for launch, mature and end-of-life phases, the arithmetic that turns scores into a ranking, and the events that should override the ranking entirely.

Six criteria worth scoring and three that only decorate

A criterion earns its place only if two credible suppliers could reasonably score differently on it. That test removes most of the usual list. Certificate count, catalogue breadth and the enthusiasm of the sales contact feel like evidence, but they separate almost nobody, and every decorative column takes weight from the ones that do. Six criteria separate display suppliers, because each maps to a failure mode with a cost attached.

Three further candidates are decoration. The number of certificates held, because a certificate names a standard and a scope, and ISO publishes the standards themselves for anyone to read; the scope line is what buyers skim. Catalogue breadth, because you are buying one part. And the speed of the sales contact before an order exists, because that clock is not the one that runs after the award.

Why the weights shift between launch and end of life

Weights are a statement about risk, so they have to be rewritten when the risk moves. In the first months the exposure is design: a module that cannot be built to print, or that fails the first thermal cycle, costs more to recover than any price difference between candidates. Once the design is frozen, exposure shifts to repeatability and price. In the final year it shifts again, to whether the part can still be built and whether the supplier will say so early enough to plan a last-time buy.

The table is a starting position, not a rule. A medical programme with a long service obligation might hold continuity at 35 percent from the first shipment, while a nine-month product may never leave the launch column. Write the weighting down before scoring, date it, and change it only at a stage gate, with the reason recorded beside it.

A high quality score cannot rescue an end-of-life part

Consider a supplier that scores five on process control, has shipped for four years without a defect traced to its line, and answers technical questions inside a day. In the fifth year the panel at the centre of your module reaches end of life, and the notice arrives ninety days before the final shipment. The quality score is accurate and, at that point, worth nothing, because there is nothing left for it to describe. The programme now needs a redesign, a fresh qualification cycle and possibly a new enclosure.

This is why continuity and change control carry weight in the end-of-life column that they do not carry at launch. The questions are concrete. Does the supplier state a longevity position for the panel and for the driver IC? Does it commit to a notice window with dates rather than a policy statement? Does it offer a last-time-buy quantity and hold it? Does it disclose sub-assembly obsolescence as well as the glass? A supplier that will not answer those in writing is not neutral on continuity; it is a risk that a high quality score is flattering.

Scoring responsiveness with a clock rather than an impression

Responsiveness is the criterion most often scored from memory, and memory rewards the account manager rather than the organisation. Replace the impression with a log. Every inbound question, quotation request, change notice and corrective action gets two timestamps: when it was sent, and when a substantive answer arrived.

An acknowledgement does not stop the clock. Score the median rather than the average, in business days with both calendars applied, because a silence that spans a national holiday is not the same as a week of silence. One anchor scale is enough: five for a complete technical answer inside one business day, three for two to three days, one for more than a week or for an answer that only arrives after escalation. Keep the raw log with the score; when a supplier disputes the number, the log settles it.

Normalising price before it enters the weighted score

Quoted prices are not comparable until they describe the same transaction. The first correction is the trade term. An ex-works price and a delivered duty paid price differ by freight, insurance, clearance and duty, so comparing them directly is a category error rather than a rounding issue. The ICC defines the terms in the Incoterms 2020 rules, and the duty rate for the destination is a separate lookup that Access2Markets supports for EU imports.

The second correction is volume and one-off cost. Tooling, NRE, test fixture development and first-article inspection belong in the comparison, amortised over the volume you actually expect rather than the volume printed in the quotation. Divide total programme cost by good units expected, not units ordered. A supplier with a higher unit price and a stable yield can beat a cheaper one whose first-pass yield forces rework and replacement shipments.

Payment terms and minimum order quantity carry a cost too, and both belong in the same line. The custom sourcing guide covers where these commercial terms are settled in the development sequence; the scorecard only consumes the result.

Running the weighted matrix and reading the gaps

The arithmetic is deliberately dull so that the argument stays in the weights. Score each criterion from one to five against a written anchor, multiply by the weight for the current life stage, and add the results. Five points are enough; ten invite false precision without improving the decision.

Weighted score = sum of (criterion score / 5) x criterion weight

Read the columns rather than the totals. Supplier B leads by five points, and the lead sits almost entirely in price and quality, while A holds the criteria that govern what happens when something changes. Re-weight the same scores on the end-of-life column and the ranking reverses: A at 73, B at 60. Neither result is wrong. The weights are the decision; the scores are only evidence.

Where a scorecard misleads: audit reports and first-lot samples

Two pieces of evidence feel conclusive and are not. An audit report describes a system on the day of the visit, against a standard that may say little about displays; it records that procedures exist and that calibration labels are current, not whether a deviation was escalated last month. A certificate is a scope statement, and the scope is usually narrower than the buyer assumes. CDTech publishes its own quality and certification documentation in that form.

First-lot samples fail in the opposite direction. They are the most careful units the supplier will ever build: engineering attention, selected components, sometimes hand-carried to the meeting. Approving a first lot is a gate, not a score. What discriminates is lot-to-lot data over time: outgoing inspection results, capability on key parameters, and a defect history that includes the unremarkable months.

Automotive programmes raise the bar on both counts, which is why their quality planning language comes from AIAG rather than from a generic checklist. In the EU, the operator duties under the General Product Safety Regulation (EU) 2023/988 put traceability and technical documentation on the same footing as the goods.

Thresholds that trigger re-evaluation or disqualification

A weighted score ranks suppliers; it does not stop one. Some events are not a matter of degree, and averaging them into a total is how a disqualifying fact gets buried under a good price. Write the triggers before scoring starts, so that invoking one is a procedure rather than an argument about a favoured supplier.

• An unannounced change to panel, driver IC or firmware: score change control at one and requalify before the next shipment.
• Two consecutive lots outside an agreed capability limit: hold incoming stock, escalate, and keep quality at two until four clean lots.
• A doubling of median response time across two quarters: re-score response and reopen the second-source question.
• An end-of-life notice shorter than your own requalification lead time: disqualify unless a last-time buy covers the remaining programme.
• A declaration missing, expired or narrowed in scope: hold the part until the paperwork matches the delivery, since a statement written under an earlier RoHS 2011/65/EU revision is not evidence for today's goods.
• Silence on longevity: score continuity at one rather than treating an unanswered question as neutral.

Each trigger maps to a criterion already in the matrix. Where a trigger and a high total disagree, the trigger wins: the total describes the past, and the trigger describes the next shipment.

Keeping the scorecard alive between annual reviews

An annual review is too slow for a criterion that can change in a quarter. The scorecard behaves better as a register than as a report: one owner, one sheet, dated entries, and an evidence file holding the response log, the lot data and the correspondence behind each score. Come review time it is a reading of the register rather than a reconstruction from memory.

Two habits keep it current. Re-weight at stage gates and nowhere else, meaning launch exit, volume ramp and the start of the end-of-life planning window, because changing weights mid-phase is how a favoured supplier gets rescued. And attach every score to a dated artefact, since a score with no artefact is an opinion, and opinions are what the scorecard was built to replace.

The register is written by the sourcing team rather than the supplier, and it is checked against each candidate's public description of its own scope before the first round; CDTech describes its company background in that form. The comparison costs an hour and prevents the most common scoring error, which is crediting a supplier with capability it never claimed.