Ask a sales team how satisfied the customers are and you get a confident answer. Ask for the evidence and you get the last visit report. In industrial business that is a risky way to run a relationship, because a buyer who is unhappy rarely says so. He moves the next program to another supplier and tells you afterward.
A supplier of machined hydraulic manifold blocks and valve housings, 320 employees and EUR 60 million of revenue, decided to measure instead of assume. The company has 86 active customers. Twelve of them carry 68 percent of the revenue. Before writing a single question, the team agreed what it would measure and why.
Three sources, three blind spots
What customers say, what they intend to do and what your delivery data shows are three views of the same relationship. Each one answers a different question, and each one is blind somewhere.
Satisfaction ratings per criterion show where customers are unhappy and how many of them. They do not show how much the topic matters to those customers. A recommendation question on a scale from 0 to 10 gives one number that is easy to follow over the waves, but it carries no reason, and with a few dozen answers it moves by chance. Delivery and complaint data are complete, monthly and free of opinion, but they are measured with your definition, not with the one your customer uses. Supplier scorecards that large customers send close that last gap, and only for the customers who send one.
The rule the team wrote down: no source is reported alone. Where two of them disagree, the definition usually differs, and that difference is itself a finding.
What the first wave showed
Fifty contacts at 31 customers answered. Overall satisfaction came out at 3.5 of 5, the recommendation score at plus 6 points. Four beliefs did not survive the first reading.
| What the plant believed | What the answers showed |
|---|---|
| Customers mainly care about price | Price named among the three most important topics by 24 percent, delivery reliability by 82 percent |
| 87.4 percent on time is acceptable | Delivery reliability rated 3.1 of 5, and the nine customers with a scorecard measured 82.7 percent |
| Quality is strong, so complaints are fine | Product quality rated 4.1, complaint handling rated 2.8, the lowest value of all |
| Our buyers are satisfied | Purchasing rated 3.8 overall, quality contacts at the same customers rated 3.0 |
The gap between 87.4 and 82.7 percent was not a measurement error. The plant counted a shipment as on time when it left the plant on the confirmed date. Several customers counted it against the date they had requested, and booked it when it arrived at their dock.
Decide the decisions first
The most useful page of the whole exercise is written before any question: why we measure, which decisions the results will feed, who owns the measurement, how often we ask, and what customers will hear afterward. The example company named three decisions: where to improve first, which accounts need attention from the management, and what goes into the departmental targets of the next year.
If nobody in the room can name a decision that will be different afterward, the survey is not worth the time it costs your customers. Quality management standards such as ISO 9001 expect you to monitor how customers perceive you. They leave the method entirely to you, which means the purpose has to come from the business, not from the standard.
One more point belongs on that page. A survey handles the personal data of your contacts. Have the contact list, the invitation and the storage of the answers checked against the rules that apply to you before anything is sent.