Key takeaways
- A rating collected under pressure measures the pressure, not the customer.
- Never assign a metric to someone who can move the number but not the underlying quality.
- Improvement never gets audited. That asymmetry is where every gamed metric survives.
- Compare your satisfaction score with your repeat rate. If they disagree, one of them is fiction.
- Read the free-text comments. Stars can be coerced at a doorstep. Sentences cannot.
Quick answer
A customer rating is only a measurement if the person giving it has experienced the thing being rated. When ratings are collected under pressure at the point of delivery, the number stops describing the customer and starts describing the pressure. This is a textbook incentive failure: a metric assigned to a person who cannot influence the underlying quality, only the number. The fix is not better training at the front line. It is moving the target off the individual and onto something they can actually control.
A delivery agent asked me to rate my food before he handed it to me.
Not after. Before. The bag was still in his hand. “Sir, please give the rating first.”
I told him I had not tried it yet. He said his manager had asked him to take a screenshot, and that he would be scolded otherwise. Then he took my phone, selected five stars across every parameter himself, screenshotted the result, submitted the review, screenshotted it again, and sent both images to a WhatsApp number.
I saw the chat. It was full of screenshots from other customers.
I was not angry with him. He looked like a person trying to complete a KPI. That is the whole story, and it is worth sitting with, because somewhere in that organisation is a dashboard showing something like 4.9 out of 5, and someone has probably been congratulated for it.
What actually broke here
This is not a story about one badly managed delivery fleet. It is the cleanest illustration I have seen of a specific management failure, and it happens in marketing teams constantly.
The failure is assigning a metric to someone who can influence the number but not the thing the number is supposed to measure.
A delivery agent cannot control whether the food is good. He cannot control whether it arrived hot, whether the kitchen was slow, whether the packaging leaked. He controls exactly one variable in the rating: whether the customer taps five stars in front of him.
So that is the variable he optimises. He is not being dishonest. He is being rational inside the system he was handed, and if you built the system, the behaviour is your output, not his.
Why leadership does not catch it
Because the number moves in the right direction, and numbers moving in the right direction do not get investigated.
I have never seen a leadership team commission an audit into why a score improved. Improvement is treated as confirmation. Decline is treated as a problem. That asymmetry is where every gamed metric lives, and it survives for years inside otherwise well-run companies.
The tell is always the same: a customer satisfaction number that is high, stable, and completely disconnected from repeat purchase, churn or complaint volume. If your rating says 4.9 and your repeat rate says something else entirely, one of those two numbers is measuring the customer and the other one is measuring your process.
What I would do instead
Move the metric to where the control is. The delivery agent can be measured on time to doorstep, condition of the package, and courtesy. All three are within his control and all three genuinely matter. Food quality belongs to the kitchen. Rate the parts to the people who own them.
Collect the rating after the experience, not during it. A prompt a few hours later gets a lower response rate and a far more honest score. Lower response rate on an honest metric beats full coverage on a fictional one, every time.
Read the free text, not the stars. Stars are easy to coerce and easy to game. Sentences are not. Nobody stands at your door and dictates a paragraph. The written comments in a review set are worth more than the aggregate score, and almost nobody reads them systematically.
Audit the good news once a quarter. Pick the metric that has improved the most and ask how it could be gamed. Not as an accusation. As a design review. If you find that it could be gamed easily and there is pressure attached to it, assume it is being gamed, because it is.
The operator move: pick the metric that has improved most this quarter and spend twenty minutes asking how it could be gamed. Not as an accusation, as a design review. You audit your worst number by instinct. Nobody audits the best one.
The uncomfortable part
Somebody set that screenshot target. Somebody built a dashboard on top of it. And a whole chain of people between the delivery agent and the leadership team looked at a rising customer satisfaction score and chose not to ask the obvious question.
That is a design problem, not a front-line ethics problem. And design problems always surface at the front line, wearing somebody else’s uniform.
Who this applies to
Anyone who owns a customer-facing metric with an incentive attached: marketing leaders, operations leaders, founders, and in particular anyone whose satisfaction score has been comfortably high for long enough that nobody questions it. My challenge to you: audit your best-performing number this quarter, not your worst.
Frequently asked questions
Why do customer ratings get gamed?
Because the person collecting them is usually measured on the score itself while having no control over the quality behind it. The number is the only variable they can move.
How do I know if my satisfaction metric is real?
Compare it with behaviour. Repeat purchase, churn and complaint volume do not lie. A high score sitting next to weak repeat rates is a measurement problem.
When should customer feedback be collected?
After the experience is complete, not at the point of handover. Expect a lower response rate and a much more honest result.
Whose fault is a gamed metric?
The system designer’s. Rational behaviour inside a badly designed incentive is an output of the design, not a failure of the individual.
Leave a Reply