In their book “The Balanced Scorecard” K&N introduces a “balanced” system of measures, but they didn’t dedicate attention to the concept of “control” of these measures. .
In this post I will explore one of the possible relations between balanced scorecard measures and statistical control charts.
The importance of statistical controlled process in a balanced scorecard.
In chapter 11, K&N wrote: ” Management processes built around the strategy articulated in the Balanced Scorecard must provide regular opportunities for double-loop learning—by collecting data about the strategy, testing the strategy, reflecting on
whether the strategy is still appropriate in light of recent developments, and soliciting ideas throughout the organization about new strategic opportunities and directions. The cause-and-effect relationships embodied in a Balanced Scorecard enable executives to establish short-term targets that reflect their best forecast about the lags and impacts between changes in performance drivers and the associated changes in one or more outcome measures”.
Well, you cannot evaluate if you make a good forecast based on a measure of the current process if you are not sure that the underline process is in a statistical control.
Let’s take as an example one of the financial and related customer measure of the balanced scored of BIT (Fig. 1).
The managers of BIT believe that it is necessary to limit the request of liquided damaged to a percentage of 0.5% of the order inflow in order to appear attractive to shareholders. They believe that there is casual-relationship effect between the request of liquidate damage and the performances of KPI tender. In other words, they think that if the organization is able to reach at least 90% of the tender KPI the request of liquidate damages will be lower than 0.5% of the order inflow. They know that for “extraordinary events” this might be not true but if the organization behaves “normally” this cause-effect link will be true.
A typical “extraordinary event” is the performance of KPI in the first six months of activity; it is accepted by the contractee not to claim liquidate damages during the first six months even if SLA are not respected.
In general, the management of BIT (and the contractee) should be able to identify when the cause-effect is not working because of an “extraordinary event” (or, in other words, that they system is not behaving in a “normal” state).
K&N suggest correlation analysis as possible tool to quantify the relationship between a financial and a customer perspective measure. Statistically, it is possible to make inference of cause-effects (and forecasts) between measures only under an assumption of stationarity. (the concept of stationarity process and the concept of statistical controlled process are strongly related).
Control Charts as diagnostic measures.
The statistical control charts offers a powerful and easy tool to understand if the assumption of stationarity (people working in the statistical control process prefer to talk about controlled process); quoting Keller “Statistical process control is synonymous with stability and is achieved when the short-term variation provides a good model (or estimate, or prediction) of the longer-term variation”.
Always quoting Keller “A key value of the control chart is to identify the occurrence of special causes so that the process knowledge can be gained concerning the sources of special causes and corresponding corrective and preventative action taken for process improvement”. For this reason, the control charts can be a good tool to use in the continous improvement process of any QMS.
In their book, K&N don’t pay attention to the concept of statistical control of their measures even when they introduced the so called “diagnostic measures” (those measure that monitor whether the business remains in control and can signal when unusual events are occurring that require immediate attention) .
K&N refer to diagnostic measures as a way to balance strategic measures to avoid bad things; in this sense, their concept of diagnostic measure is far away from the concept of statistical control. However, I like the analogy that they make between diagnostic measures and a thermometer. “We don’t devote enormous energy to optimize our body temperature” they say. I would add: “we have not to know the thermal equilibrium and the zero law of thermodynamics to read a thermometer”. I would also add: “we have not to be either a doctor or a nurse to be able to use a thermometer”. What we must absolutely know to read a thermometer is just the level at which we might have fever; on the other part, we must have also some basic information of how to use the tool (ex. in which hole of the body to insert it, not to expose before or after the use to high temperatures).
I wish to make the same analogy with the statistical control charts. You have not been a disciple of Deming or to know the central limit theorem to read and use a control chart. You just need to know to read when “you have fever” and know some basic precautions to manage the tool.
I will illustrate in further posts how BIT uses specific control charts to monitor if the process measured by a balanced scorecard measure is under control. Here, I just want to introduce the basic reading of a control chart. The measure must be included between an upper and lower limit (in analogy with the “thermometer” the normal temperature of the body should be included between the upper limit of 37.5 and the lower limit of 36.5 degree).
Like it is important to check that thermometer has not been stored in a warm or cold place before using it, it is as well important to check some basic conditions of the measures before using the control chart; I will talk about these precautions of usage when I will introduce each specific control chart, starting from the p-chart.