There is a failure mode particular to businesses that take customer feedback seriously: they act on all of it. A detailed, articulate, deeply unfair review lands on a Tuesday, and by Friday a working process has been redesigned around a single person's bad evening.
The discipline that prevents this is not scepticism. It is four questions, asked in order, before anything changes.
1. Is this about a process or a person?
A complaint about a process — the booking form, the waiting, the handover, the invoice — describes something that will happen again to somebody else next week. A complaint about a person describes something that may never recur, and that requires an entirely different conversation held privately.
Businesses routinely treat process complaints as staffing problems, which is expensive and unfair, and staffing problems as process problems, which changes nothing.
2. Has it happened before?
Go and look. Not from memory — memory over-weights whatever arrived most recently and most loudly. Search your own public feedback for the specific words, across every platform, over at least a year.
One occurrence is an anecdote. Two starts to be a question. The threshold matters more than it sounds, because the difference between acting at one and acting at two is the difference between a team that improves and a team permanently reorganising itself.
3. Where else did it appear?
The same theme on two or three unrelated platforms is a much stronger fact than the same theme three times in one place. Independent sources cannot coordinate, and a pile-on can. This is the test that most reliably separates a real operational problem from a bad fortnight or a campaign.
4. Is it getting worse?
A theme with a long tail and nothing recent is usually a problem you already fixed. A theme that was quiet and is now arriving weekly is the one that deserves this quarter. Direction matters more than volume, and it is the dimension most often ignored because it requires having read the same sources before.
Direction beats volume. A loud old complaint is history; a quiet new one is your next quarter.
Reading a theme over time
What the four questions are protecting you from
Two failure modes, and a team can only avoid one at a time without a rule. The first is inertia: a genuine, recurring, worsening problem that everybody has read about and nobody has owned, because each individual instance looked small. The second is thrash: a process rebuilt every few weeks in response to whatever arrived most recently, which exhausts the people doing the work and improves nothing.
Thrash is the more expensive of the two and the less discussed, because it looks like responsiveness. A team that has changed the handover procedure three times this quarter is not a responsive team. It is a team without a threshold.
Write the threshold down before you need it
The single most useful thing a service business can do with customer feedback is decide, in advance and in writing, what will trigger a change. Deciding in the moment means deciding under the influence of whichever review is currently on the screen.
- A theme named twice or more in a fortnight, or on two or more unrelated platforms, gets looked at.
- A theme that is worsening gets an owner and a date, whatever its absolute volume.
- Anything below that gets recorded and left, which is a decision rather than an omission — and recording it is what makes the second occurrence visible.
- Nothing changes on the strength of one review, however articulate, however senior the person forwarding it.
That last line is the hard one, and it is why the threshold has to exist on paper before the review that tests it arrives.
And then the fifth question, which is about money
Once a theme passes all four, the last question is what it costs to leave alone versus what it costs to fix. Not every genuine problem is worth solving. A recurring complaint about parking at a site you do not control is real, confirmed, worsening, and still not your next project.
What makes this answerable rather than a matter of opinion is having the frequency, the spread and the trend in front of you as numbers somebody counted. That is the entire argument for putting this work on a cadence instead of doing it in response to whatever arrived this morning.
What to take away
- Separate process complaints from person complaints before anything else; they need opposite responses.
- Check recurrence in the record, not in your memory.
- The same theme on unrelated platforms is far stronger evidence than repetition in one place.
- Direction — is it worsening — matters more than raw volume.
- Then weigh cost of fixing against cost of leaving. Some real problems are correctly ignored.