Complaints Are Your Best Process Diagnostic — and the Only One You've Already Paid For

The symptom

Almost every business has a complaints folder, a complaints log or a field in the ERP. It gets maintained because a standard or a customer requires it. It rarely gets analysed, and when it does, it gets analysed as a rate: "We're at 1.4 per cent, last year it was 1.6."

That number answers whether things are getting better or worse. It does not answer the only question that would change anything: where exactly in the process does the error originate that the customer eventually sees?

The answer is usually already sitting in the existing records. It has simply never been sorted in a way that lets you see it.

The mechanism

A complaint is an error that survived every internal check. That is what makes it diagnostically valuable: it doesn't mark just any error, it marks precisely the kind your process is blind to.

The usual handling destroys that value in two ways.

First, categorisation is by symptom, not by cause. The typical categories read "wrong delivery", "damage", "quantity", "quality". That describes what the customer noticed. It says nothing about where it arose. A wrong delivery can originate in order entry, in picking, in master data maintenance or in dispatch — four entirely different processes, one category.

Second, the individual case gets solved, not the pattern. Complaint handling is oriented toward customer satisfaction, and rightly so. The case is clarified, replaced, credited — and thereby closed. What's missing is the second look across all cases, which reveals that eleven of the last thirty share the same point of origin.

There's also an uncomfortable dark figure: the most serious process errors frequently produce no complaint at all. The customer simply stops ordering. This is why complaint rate is a weak measure of satisfaction — as a process map, it remains strong, because it shows where your system is permeable.

The cost

An anonymised example, altered in detail, real in pattern. A manufacturer of plastic components had held a stable complaint rate of around 1.8 per cent for two years, unremarkable against its sector. An improvement project had been discussed several times and deferred for lack of a clear starting point.

We then analysed 214 documented complaints from 24 months — not by symptom, but by reconstructed point of origin. That took about eight working days and required no new data collection.

The result was sharper than expected. Around 40 per cent of all cases traced back to two causes. The first: tool changes on one machine group without a documented first-article release, producing dimensional deviations across a run that were only noticed at the customer. The second: articles with multiple revision levels, where picking could not distinguish the level from the article number alone.

Both causes were known in the business — as an annoyance, not as the main item. Nobody had added them up.

The direct cost of that 40 per cent came to roughly €96,000 over two years: replacement deliveries, freight, credit notes, internal handling. Not included: the customer who declined to renew a framework contract after two incidents.

The fix

Both causes could be addressed without investment.

For tool changes, a mandatory first-article release was introduced — with one condition that made the difference: the gauge sits at the machine, not in the inspection room. A rule requiring an 80-metre walk gets circumvented under time pressure; one requiring two metres doesn't.

For revision levels, picking got a visible distinction: separate storage locations per revision and a coloured label. Not a system change — a sorting.

After twelve months the complaint rate had fallen to 1.1 per cent. The distribution was more interesting: the two main causes had essentially disappeared, while the number of one-off cases was unchanged. That is what solving a pattern rather than symptoms looks like.

This analysis requires no new data collection, only a different sorting of data you already hold — and that is exactly where a process analysis starts. Fixed scope, fixed duration, fixed price.

The next step

Take your last thirty complaints and write beside each one not the symptom, but the process step where the error originated. If more than a third concentrate on two steps, you've found your next project — and it costs less than you think.