I was looking at hours of work sold per car in a dealership analysis. One person’s average had fallen from about 2.89 to 1.65 between the first and sixth reporting periods.
Roughly a 43% drop. My first reading was that their performance had fallen badly.
I could see how that number might become a conversation about effort or selling ability. A manager would have a clear figure to point to, and the person being questioned would have something large to explain.
Then I put the average beside the numbers used to calculate it.
The part I almost missed
| Measure | First period | Sixth period |
|---|---|---|
| Hours sold | 207.9 | 206.1 |
| Cars | 72 | 125 |
| Hours sold per car | 2.89 | 1.65 |
The total hours of work sold had barely moved. It was down by less than 1%. The number of cars had gone from 72 to 125, an increase of about 74%.
The person was handling many more cars while selling almost the same total hours of work. The average fell because those hours were spread across a much larger number of cars.
The calculation was correct. The story I had started building around it was too confident.
I kept the average in the analysis. Less work was being sold per car, and that change still needed explaining. But I withdrew the interpretation that someone’s work had collapsed.
What the table couldn’t tell me
I didn’t know how those cars had been assigned. I didn’t know whether they needed comparable amounts of service. The table couldn’t tell me how much say this person had in the work they received.
Those gaps left room for several explanations. Different service needs could matter. So could other things. I couldn’t pick one just because it made an easy story.
I also couldn’t swing to the opposite conclusion and declare that the person had performed brilliantly. Handling more cars doesn’t settle every question either. I needed to be as careful with a reassuring explanation as I had been careless with the first one.
That’s the uncomfortable part of reading numbers about people. The calculation can be finished while the fair judgment is still a long way off. Putting a percentage beside a name doesn’t tell me which part of the result was within that person’s control.
The advice I could stand behind
I changed the recommendation to look at the assignment of cars and the work involved before deciding what coaching was appropriate.
That was a change in my analysis. I don’t have a measured business result to attach to it, and I don’t know what a closer look at the assignments would have found.
What I do know is that the first interpretation could have sent someone into a conversation they weren’t prepared to have fairly. The missing information would have become the employee’s problem to explain after the report had already made its impression.
I’ve kept the average, the totals and the withdrawn interpretation together. I want another reader to be able to see the step where my thinking changed.
Before I could recommend telling someone to do better, I needed to understand what work they’d actually been given.