Consumer surveys draw a random sample from a large population. A supplier with 86 customers has no population to sample from. It has a revenue structure, and the first decision is whose voice counts.
Classify, then choose by rule
The example company sorted its customers by revenue of the last twelve months. Twelve class A customers carry EUR 40.8 million, 26 class B customers carry EUR 14.4 million, and 48 class C customers share the remaining EUR 4.8 million. Sales then marked the customers with high potential, whatever their size today.
The selection rule was written in three lines, which matters more than the lines themselves, because a rule can be repeated in the next wave and defended in the next management meeting.
| Class | Rule | Customers selected | Contacts each |
|---|---|---|---|
| A | All customers, four functions | 12 | 4 |
| B | The 17 largest and three with high potential, two functions | 20 | 2 |
| C | Only customers with high potential, one function | 12 | 1 |
Forty-four customers, 100 invited contacts, EUR 54.8 million of revenue covered. That is 91 percent of the business, which is the number the management cared about when it approved the plan.
Two groups are easy to lose. Customers with falling volume or an open dispute belong in the survey, not outside it. Sales proposed to skip one of them because the timing was bad. The timing is never good, and that customer turned out to be one of the seven accounts at risk. Lost customers are a separate exercise: three questions on the phone, not a questionnaire.
One contact is not the customer
The strongest finding of the first wave had nothing to do with the questions. Purchasing contacts rated the supplier 3.8 overall. Quality contacts at the same customers rated it 3.0. The account managers mostly talk to purchasing, and they had reported good relationships everywhere.
Four functions see four different suppliers:
- Purchasing judges price, quotations, contracts and the relationship as a whole.
- Quality judges defects, complaints and how fast and how well you answer.
- Logistics and scheduling judge dates, quantities, delay notices, packaging and paperwork.
- Engineering judges technical support, drawings, samples and changes.
For class A customers the example asked all four. For class B, purchasing and quality. Where your parts go into a specific plant, add the scheduling or production contact there, because the person who lives with your delivery performance is not always the person who signs the order.
The account managers named the contacts and the owner of the measurement checked the list. A list that contains only friendly contacts is worthless. The names are personal data, so the list, the invitation and the storage of the answers were checked against the data protection rules before the first email went out.
What the selection decides
Response rates differ by class, and so do the results, which is why the selection is recorded with the answers. In the first wave 26 of 48 class A contacts answered, 18 of 40 in class B and 6 of 12 in class C: 50 of 100 in total, from 31 of the 44 selected customers.
The thirteen customers who gave no answer at all are not a gap in the data. They are a finding. Each of them received a call from its account manager, and three of those calls produced more than the questionnaire would have.
Keep the rule and the contact structure unchanged in the next wave. A result that moves because you asked different people is not a result.