My Best Operators Had The Worst OEE

My best operators had the worst OEE. Not because they were bad, but because they were the only ones who could run the hard jobs. OEE was the metric. On paper, my strongest people looked like my weakest.

They knew it, I knew it and it made performance discussions impossible.

We had over 200 dies in that department. Every SKU had its own die, some were easy to run and some were not. 

And more than that, OEE didn't tell me what to do. It was a number that described a problem without pointing at a solution.

So I tried something different.

I created a database that calculated the true departmental average speed for every single die. Then I measured each operator not by raw output, but by how many hours of average production they generated. 

This normalized the die assignment so that we could compare operator production to department average.

It leveled the playing field. Suddenly the rankings matched reality, my best operators

rose to the top, my weaker ones fell to the bottom.

The operators dubbed this metric: "The Stig" (if you've seen Top Gear, you know why).

Here's what I learned from "The Stig".

One operator was at the bottom of that list. I couldn't reconcile it: he was diligent, intelligent, and clearly cared.  

So I broke out his individual metrics against departmental average, product by product.

He was above average on almost everything.

Except feeder stops.

That single issue was tanking his entire performance. Once I knew that, I knew exactly what to fix. We learned how he was doing it and we found the issue.

We fixed it.

But it didn't stop there.

I also analyzed jobs by who set them up. This same operator, the one with the worst overall numbers, produced the highest-performing setups in the department. Even with his own feeder stop issue dragging his jobs down, if he set the job up, it outperformed everyone else's on average.

We all had to learn from him.  

Here's the point I want to make to every Ops leader reading this:

Your metrics should simplify decisions, not just describe problems.

If you're running on OEE and downtime and you're not sure what to do about them, that's not a performance problem. That's a metrics design problem.

Your metrics should obviate decisions.  

You make decisions in the dimensions of the fishbone diagram: people, methods, materials, machines... etc..

Your metrics should show you outliers in every category.  Knowing an operator who's an outlier on top is just as important as knowing the one on bottom.

The top level of each metric should answer the question: who do I learn from, and who needs to learn from me?

If your current dashboard can't answer that question clearly, you're not getting what you need from your data.

Shape your metrics around the decisions you need to make. And if you're not sure how to do that I'm happy to discuss. 

Have you ever looked at a report and thought, "That can't be right"?