KPIs nobody reads, and the one that changes everything

I see three causes that repeat themselves in plants. The first: the KPI is too far away from the action. A monthly delivery performance for the whole plant tells the shift leader nothing about what he should do differently today. He cannot influence it, so he ignores it.

The second: the KPI is used as a verdict, not as a tool. If a poor OEE in the meeting leads to the question of who is to blame, people quickly learn to make the number look better instead of solving the problem. Stoppages are booked as changeover time, rework runs outside the count. The KPI rises, the plant does not get better.

The third: no action follows. The number is read off, commented on, ticked off. Nobody asks what we are doing about it today. Without that question, every KPI is just a rear-view mirror. It shows what was, and changes nothing about what is coming.

Overall equipment effectiveness is made up of availability, performance and quality, multiplied together. That is generally known and quickly explained. The trap lies not in the formula but in the use.

OEE is an equipment KPI. It says how well a particular machine uses its planned time. It says nothing about whether that machine should be running at all. I have seen plants that clung to a high OEE and in doing so produced stock that nobody needed, because the machine was not allowed to stand still. That is overproduction, and overproduction is the worst waste, because it hides all the others.

This is how I use OEE so that it changes behaviour: only at the bottleneck, not everywhere. Broken down by loss type, so that the team can see whether changeovers, breakdowns or minor stoppages are the problem. And discussed daily with the question: which loss was the biggest yesterday, and what are we doing about it today? An OEE that appears as a single bar in the monthly report moves nothing. An OEE that is broken down into its losses at the board in the morning is a work order.

Lead time measures how long an order takes from receipt to delivery. For me it is the most honest KPI in the plant, because it does not only reflect the hall. A large part of the lead time arises before the first part is processed: in order clarification, in production planning, in purchasing, in release loops.

An example: in a plant with several assembly lines, plant management was convinced that assembly was too slow. When we mapped the value stream from customer order to delivery, it turned out that the order spent the largest part of its time in order processing and in engineering release. Assembly was only a short section at the end. Until then, all improvement projects had targeted that short section.

For lead time to change behaviour, it has to be broken down into sections that someone is responsible for. Order processing needs its own lead time, production planning its own, manufacturing its own. Then each team can see where its orders are sitting, and then a board in the clerical office helps just as much as one on the line.

Delivery performance is the KPI the customer feels. It answers the question of whether we keep what we promised. And that is exactly where its catch lies: the promise is often the problem, not the delivery.

I know plants in which sales promises dates that manufacturing can never meet, because nobody asked beforehand. Delivery performance falls, manufacturing is criticised, and the actual error lies at an interface that appears in no KPI. Anyone who wants to improve delivery performance therefore first has to look at the date promise: who promises, on what basis, with what feedback from planning?

For delivery performance to change behaviour, I recommend two things. First: look at every missed date individually, not just the rate. The causes are usually few and they repeat. Second: put the KPI up at the interfaces where dates are created, in other words between sales, planning and manufacturing, and not only at dispatch. Where the date is made, it also has to be visible.

Whatever KPI hangs on the board, there is one question that makes the difference: what are we doing about it today? If that question is not asked in the shopfloor meeting, the meeting is a reporting round. If it is asked, it is a steering instrument.

For the question to be answerable, the teams need three things:

• The KPI has to sit at their level. Not the plant's delivery performance, but the on-time delivery of their orders this week.
• They have to be able to influence the KPI. If the loss arises from missing parts, the team needs a way to escalate that to purchasing, and an answer that comes back.
• They have to be allowed to act. A team that has to get every small change approved first stops making suggestions.

The third is a leadership question, not a KPI question. And this is the reason why I never introduce KPIs without leadership work. The best board is useless if the supervisor is not allowed to decide.

Finally, a piece of advice that is uncomfortable: put up less. I see boards with twelve charts of which nobody looks at more than two. Every KPI that does not lead to an action dilutes the attention for those that do.

Choose a few KPIs per team that it understands, can influence and discusses daily. Check every few months whether they still trigger actions. If not, swap them or drop them. A KPI is not a monument. It is a tool, and tools that nobody uses do not belong on the wall.