A table of eight means tells you almost nothing. Two criteria with the same average can hide completely different pictures, and a criterion nobody cares about can look like a problem. The first wave of the example supplier was read with three columns side by side.
| Criterion | Mean | Rated 1 to 2 | Named in top three |
|---|---|---|---|
| Delivery reliability | 3.1 | 26 % | 82 % |
| Product quality | 4.1 | 0 % | 72 % |
| Complaint handling | 2.8 | 43 % | 44 % |
| Lead time | 3.4 | 18 % | 28 % |
| Information and communication | 3.0 | 38 % | 26 % |
| Price for performance | 3.3 | 20 % | 24 % |
| Technical support | 4.3 | 3 % | 16 % |
| Quotation speed | 3.2 | 28 % | 8 % |
The average of the eight means is 3.4, and the average share named is 37.5 percent. Those two lines split the table into four fields.
Four fields, two of them for action
Act first holds what is named often and rated below average: delivery reliability and complaint handling. These get the resources, and two fields are the maximum, because a list of eight priorities is a list of none.
Keep up holds product quality, named by 72 percent and rated 4.1. This is what customers buy you for. A drop here costs more than any gain elsewhere.
Watch holds information, price and quotation speed. It is not a waste bin, as the next section shows.
Hold holds technical support and lead time, where extra effort is not noticed.
Sales had expected price to come out on top. Three of four contacts did not name it among their three most important topics, and it sits in the middle of the ratings. The customers of this plant do not leave because of price. They leave because parts are late and nobody tells them.
The hidden driver
Stated importance is what people say matters. The driver is the link between a rating and overall satisfaction, measured as a correlation from minus 1 to plus 1. The example calculated it per criterion:
| Criterion | Named in top three | Driver |
|---|---|---|
| Information and communication | 26 % | 0.82 |
| Delivery reliability | 82 % | 0.75 |
| Complaint handling | 44 % | 0.67 |
| Product quality | 72 % | 0.41 |
| Price for performance | 24 % | 0.09 |
Information was named by only a quarter of the contacts, yet whoever is unhappy with it is unhappy overall. It moved to the action list, and the open answers confirmed why: almost all described late notice of a delay. A correlation shows a link, not a cause, so it earns an action only when the wording of the answers agrees.
Two criteria, the same mean, different problems
Delivery reliability came out at 3.1, information at 3.0. The distributions tell different stories. Delivery was rated 3 by most contacts. Information split the field: 19 of 50 rated it 1 or 2 and 17 rated it 4 or 5. A split means some customers are served well and others are not, and the right next step is to find out who, not to design a general improvement.
Be honest about small numbers
With 50 answers, one contact is worth two percentage points. The recommendation score of plus 6 carries a rough range of about plus or minus 20 points, calculated as two standard errors. Class C, with six answers, produced a score of plus 50 with a range of plus or minus 41, which means it is not a result and was reported as six counts instead.
The rule the team adopted: below ten answers, report counts, never percentages, and read every answer one by one. In industrial business a single answer can stand for a customer with millions in revenue.