Operational Meaning - how to put ideas to work David and Sarah Kerridge Introduction To make any idea achieve a useful result, we must give it operational meaning. This is a definite form that can be commun- icated, tested, and improved. For example, a chef has a new idea for a cake. Others could understand the idea, but not produce the same result. The operational form of the idea is a recipe which specifies the ingredients and the way to cook it. It is easy to think of other examples: an engineer plans a new type of car. The blueprints give the idea operational meaning, from which others can build a prototype, test it, and revise the blueprints. Giving ideas operational meaning is a technique that has many uses. Used correctly, it is an important tool of continual improvement. It makes learning and communication much easier, and has a direct effect in reducing variation. Used without under- standing, it can put ideas in a straight-jacket and hold up progress. Operational definitions The first, and simplest use is to construct operational defin- itions of concepts such as "time", or "size", or "a breakdown". An operational definition is not so much "true" as useful, for a particular purpose. Its value is that it states or implies the same actions to everyone who uses it. A military unit is planning a night attack. The officer in command gives the order "synchronise watches". Everyone then sets their own watch to the time shown by the watch of the leader. This provides the operational definition of time for the duration of the particular sortie. Whether it agrees with the "right" time, determined by Greenwich Observatory, does not matter. It makes sure that they keep exactly in step with each other. In the same way there is no "right" definition of a clean table, (to use Dr Deming's example), a hot dinner, or a patient's temperature. An Operational Definition must be fit for a partic- ular purpose, and result in sufficiently close agreement when used by different people, or at different times and places. Making sure of agreement in this way not only increases clarity, it improves stability. Without it, a process can appear to be out of control when it is really the observation system that is unstable. So "corrective" action is taken which actually makes things worse. Finally, it directly reduces variation. Does it suit the purpose? It is easy enough to find out whether a suggested operational definition produces good enough agreement. We can set up an experiment in which, for example, different people observe the same thing. If they record the same result, to the accuracy that matters, the agreement is satisfactory. If not, improve it using the Deming cycle, (PDSA). In the same way, we can check on consistency over time, or from place to place. By contrast, there is no easy way to decide whether an operational definition represents the original concept suffic- iently well. We must usually rely on understanding of the subject matter. Should a patient's temperature be measured on the skin, under the arm, in the mouth, or in the rectum? It depends on the purpose. If the purpose is not obvious, consider the action to be taken when the result is known. The ideal way to give operational meaning to "better for the purpose" would be to see if one suggested operational definition enables us to make more accurate prediction than another. This is far more difficult and time consuming than simply testing whether different users agree. For example, we could compare different rules for class- ifying children as dyslexic, and see which system more accurately predicts each child's reading ability in several years time. Sometimes this kind of investigation is profitable, but less often than we might think. Unless two suggested operational definitions classify a substantial proportion of cases differ- ently, the scope for improvement is small. Even then, so many other variables will affect the outcome, that an improvement will be hard to detect. Since the gain when any reasonably good operational defin- ition is introduced is so marked, and the difficulty of further refinement by a prediction study so great, other improvements usually take higher priority. Still, it is always wise to be on the look out for exceptions. Formulation of a theory A more elaborate use of operational meaning arises when we try to turn a suggestion for improvement into a form which can be tested. What often happens is that a group get together, and try to explain why things are going wrong. However convincing the various explanations sound, we must not rely on persuasive arguments, but facts. From a possible explanation, we derive a suggestion for action. Then to test this idea, we must state it as a precise prediction, that if we do A, under conditions B, we will observe C. To take a simple example, consider the idea that vitamin supplements in pregnancy lead to healthy babies. To test it we must specify which vitamins, in what dose, given for how long, to which group of pregnant women, and how we are to define a healthy baby. An example from crime prevention For a more elaborate example, police forces want to find ways of reducing crime. One possibility is that video cameras scanning the streets would be effective. This may be a good idea, but it is not in a definite enough form to be tested. We must ask a great many questions, which define the system to be used. How precisely are the cameras to be used? How many cameras? How high above the street? Are they to be visible, or concealed? Are the pictures to be continuously viewed at police head- quarters, or recorded on tape, or both? If an incident is seen, what type of action is to be taken? Is the idea to send a police patrol at once, or to identify the individuals and prosecute later? Under what conditions is the system to be tested? Is it to be tried in city centres, or residential areas? Does it work only in well-lit streets? Which types of crime do we expect to deter? For example drug- dealing, car theft, or violence? How will decide if it works? Do we mean that doing this in addition to what is done now would reduce crime. Or do we mean that, with the limited resources available, video cameras are more effective than spending the same amount in other ways? (More policemen, more frequent patrols, better street lighting, etc) How would we know if there is less crime? Do we measure incidents reported, or numbers of convictions, for example? Is it a success or a failure if the crime simply moves elsewhere? If the measure chosen shows an improvement, is this due to the publicity surrounding the project, or to the video cameras themselves? Asking so many questions sounds like a statistician making difficulties. Fortunately few problems in management are as difficult as the crime problem. The point is that an idea, in the abstract, can not be tested. We can test the prediction that a well defined system which uses video cameras will under partic- ular circumstances reduce particular measures of crime. If our test demonstrates an improvement, we have made the idea work. If not, the idea may still be a good one, but must be operationalised in a different way: for example, we could try different numbers of cameras, or use them in different types of area. This is yet another example of the Deming cycle. When a hunch or an "explanation" has been given operational meaning, it becomes a "theory", in the technical sense used in discussing scientific method. It can be used to make predictions, and if the predictions are correct, the theory is "accepted". This does not mean that we believe in it without question, but that we use it until something better is found. The next step is to look for exceptions, so that we know the range of conditions under which it applies. You do not understand a theory unless you know when it will not work. An explanation not expressed in operational terms as a prediction does not help us to learn, because it can never be proved wrong. Although the idea of operational meaning is useful and simple, it can cause problems if introduced without sufficient education. This is because anyone who does not understand why it is necessary will think we are being unreasonably particular about details, but ignoring what really matters. For example, when someone has thought hard about a problem, and come up with a creative suggestion, it seems hard-hearted and obstructive to ask so many questions, and to insist on strict definition and testing. This is particularly true in the prevail- ing business culture, which mistakes instant, decisive action for leadership. A useful approach is to say "this is such an important idea that it deserves very careful study". This is both true and tactful, and should avoid hurt feelings. Conclusion Although we must stress the need to define exactly, and to test each new idea thoroughly, we must maintain balance. One of the follies and failures of our education system in the past has been to present Arts and Humanities as the creative subjects, and the Sciences as cold and disciplined. This is nonsense: ask a ballet dancer whether discipline is needed, or a scientist about the excitement of discovery. All learning comes from a living partnership between creativity and imagination, on the one hand, and patience, caution, and discipline on the other.