menu SIGN IN
Michael Taplin
Michael Taplin is the home of The KPI Bible and offers the lowdown on KPI modeling

Corporate Dynamics Ltd
Principal Consultant


Knowing when to act – and when to leave well alone.

Posted almost 10 years ago

Modern technology has the capacity to generate huge volumes of data, leaving the human brain with the problem of making sense of it all. What is significant? How much change do we need to see before it becomes significant; before it becomes necessary to act?

If our performance indicators are real KPIs, that is, the real levers of productivity and profitability, we will observe change over time.

The questions then become:
• How much change is normal variation within the system, a fundamental characteristic of the business system?
• What defines the point where some special cause is operating, requiring a management response?

It is so easy to slip into the trap of over-reacting to small movements in performance indicators. Unnecessary management actions often create more problems than they solve because they destabilize a smooth running system leading to reduced productivity and profitability.

Fortunately there is a reliable way to resolve this problem, but we need to look beyond our KPI theory to apply statistical control theory from the field of Total Quality Management.
If we take selected KPIs and use a Control Chart to track them we will receive reliable signals about the occurrence of an event that demands management attention.

A control chart is a data series over time, with upper and lower control limits delineated.

A data point that lies outside the control limits indicates that some special cause, some error, system breakdown or event has occurred to create the deviation from the normal range of variation. If the control limits are set at twice the standard deviation (STDEV) of the data series, we can be 95% confident that a special cause is operating. If we use 3 times STDEV we can be 99% confident.

This is a reliable signal that something has changed in our business system and we need to investigate. If it is an undesirable change we need to fix the problem. If it is a good change we should look for the cause to see if it is repeatable.


Here are a few reasons to try out this idea:
• Avoid management-induced instability. If your business system is working well, in the stable state with minimal errors, your profitability is optimal, stable and predictable. Instability always degrades performance.
• Improve relationships with supervisors and teams. No-one will tell you but, management interference can be the bane of the lives of the operators in your system.
• Find the leading indicator KPIs for your business system. Get early warning of a looming problem. An example could be a change in sales lead conversion rates or sales forecast error.
• Management’s focus on the things that are strategically significant, such as reducing system variability, and improving capability can improve dramatically.

Of course, all this works best if you have used a KPI Model to select your KPIs so that you are measuring the significant levers of performance.

Michael Taplin publishes articles and newsletters on his website, where you can download a 7 page whitepaper on this topic to share with your colleagues. Check out the KPI Modeling Group for regular news on KPI model development.

Comments (0)

Log in to post comments.