Method

Quality function deployment (QFD)

In French: QFD — déploiement de la fonction qualité

QFD is a matrix that links customer expectations to the technical characteristics of the product, weighting each link to bring out the characteristics on which design must focus its effort.

The matrix and what it crosses

In rows, the customer expectations, with their relative importance. In columns, the technical characteristics on which the company can act. At each intersection, the strength of the link between the expectation and the characteristic — strong, medium, weak or none.

The weighting of the columns is then derived from the crossing: a characteristic strongly linked to several important expectations comes out on top. That is the main result, and it designates where to focus the design effort.

The complete form adds a roof, where the characteristics that contradict one another are noted. It is the most useful part and the most often omitted: it names the trade-offs before they are settled implicitly by the first person who decides.

What the exercise really produces

It forces the relative importance of the expectations to be written down, which an open discussion never does. As long as everything is important, no trade-off is possible, and the final product is an endured compromise rather than a chosen one.

It makes orphan expectations visible: a row with no cell filled in signals an expectation that nothing in the design satisfies. It is the most profitable finding of the matrix, and it appears without any calculation.

It also reveals the useless characteristics: a column with no strong link corresponds to a technical effort the customer will not perceive. Those columns are defended vigorously, because they often carry a long-standing engineering pride.

Its right scale

The full method chains several matrices, from the expectations down to the production parameters. That chain is heavy, and few organisations take it all the way — the ratio between the effort and what is drawn from it degrades markedly after the first matrix.

The first matrix, on its own, brings the essential: the hierarchy of expectations, the priority characteristics, the uncovered expectations and the contradictions. It is the format to retain for a first use.

It does, however, assume that the voice of the customer has already been collected. Filling in the rows from what the team assumes the expectations are produces a matrix that is coherent, convincing and entirely built on assumptions — the worst possible result, since it gives suppositions the appearance of methodical work.

The common mistakes

Filling in the rows with no customer collection. The matrix then lends methodical authority to suppositions, and the discussion shifts from "what does the customer want" to "what score to give".

Running the complete chain of matrices on a first project. The effort discourages the team before the end, and the method is then filed as too heavy for the organisation.

Ignoring the roof. The contradictions between characteristics exist with or without a matrix; not writing them down amounts to letting them be settled by default, at the last moment and by a single person.

On the shop floor — Household appliances

On a range overhaul, the matrix brought out a row with no strong link at all: ease of cleaning, cited by a majority of the users interviewed. No technical characteristic in the specification related to it. The subject had been known to after-sales service for years and had never reached design.

Frequently asked questions

Is QFD reserved for product design?

No, it applies just as well to a service or an internal process: the rows become the users’ expectations, the columns the levers available to the organisation. The logic remains identical — rank the expectations, measure their link with what can be acted on, and spot what is covered by nothing.

How many expectations and characteristics should be retained?

A readable matrix rarely exceeds about fifteen rows and as many columns. Beyond that, filling it in becomes an endurance exercise and the weighting loses its meaning, each cell weighing too little to change the ranking. Grouping close expectations before starting costs less than abandoning a half-filled matrix.

How is the importance of the expectations set?

From the customer collection, by counting the frequency of mention and supplementing it with a ranking asked of the customers directly — distributing a hundred points among the expectations works well. What must be avoided is scoring by the project team, which reproduces the company’s priorities under cover of representing the customer.

What is the difference between QFD and a CTQ?

The descent to critical-to-quality characteristics translates an expectation into measurable quantities with their limits, expectation by expectation. QFD works on the whole set at once and adds two things: the cross-weighting, which ranks the characteristics among themselves, and the contradictions between characteristics. The two are complementary, QFD coming in rather at the design stage.