But before I relieve you of the stress, I'm going to tell you about the three types of change processes. They have been described by Ralph Stacey, a British researcher into complexity and organizations, in his book The Chaos Frontier:
Closed change Unambiguous problems and clear connections between cause and effect are the distinguishing characteristics of closed change. People are able to accurately forecast the effects that closed changes have on a system. It is possible to control such changes in a formal and analytical way. Examples of closed change might be: changing the customer's logo in an application, watching 188 episodes of South Park, and shopping for a loaf of bread. With closed change a task has a known purpose, including any alternative ways of achieving that purpose, with a corresponding effort that can be correctly estimated or calculated.
Contained change Management of contained change takes the form of probabilistic forecasts. In this case the effect of one particular change cannot be determined accurately, but similar historical tasks and events in the past provide valuable empirical data on the probability distribution. Examples of contained change could be: changing the style and theme in an application (more complex than just a logo, but still something we have experience with), painting an entire house (if you've done such things before), and shopping for your 40th birthday party (which, in my case, is next Monday. Remember that!) With contained change the purpose of the task is still known, though the estimate of effort is somewhat blurry due to many possible changes. However, the variability of those changes is restricted to a range we understand well.
Open-ended change For open-ended situations control is impossible. The changes themselves and their consequences are unknown to us, so we cannot compare probabilities to similar historical tasks and events. There can be many reasons. We're unfamiliar with the problem. The purpose of the task may be vague. The people involved may not be not fully committed. Or the information available is insufficient or subjective. In short, the situation is unique and/or unclear, and it defies both analytical and empirical forecasts. Examples of open-ended change might be: migrating an existing web design to a new and unproven technological platform, painting the entire neighborhood (if you've never done such a thing before), and building a bridge from Denmark to Sweden.
Let me give you a personal example:
Two weeks ago I cooked a dish of which the recipe says it takes only 17 minutes to make (which, being a man alone in the house, was of course the reason I selected it in the first place). The first time I made it I was surprised that the dish took me exactly one hour to make. Either I'm a very slow cook, or the recipe was lying. (I'm settling for the latter option.) Needless to say, that first time my cooking efforts were faced with open-ended change. Everything took longer than I had anticipated. And the recipe was silent about some essential tasks. Fortunately, the result was delicious…
Today I made that same dish again. However, due to my critical spouse being in the house, I changed some of the ingredients. I swapped the canned vegetables for fresh ones that I had to cook myself. I knew that cooking these vegetables would take about ten minutes, so I added that time to my expected delivery time of one hour. It is a fine example of contained change. Both numbers (one hour plus ten minutes) were derived from earlier experiences. Short from an exploding kitchen the impact of all expected changes could simply be estimated using empirical data.
By repeatedly cooking (more or less) the same dish I am able to reduce the estimation problem from open-ended change to contained change. However, it will never be a closed change problem, as it is impossible to calculate how long it takes to cook a dish. Empirical data is all we have.
The three types of change (closed change, contained change, and open-ended change) each require a different approach to estimation. They are not three discrete categories, but rather a scale from closed to open-ended. Depending on the information available, problems can move in either direction:
Problem size, complexity, and uncertainty will push a problem to the right side of the scale. Reliable information and communication can pull a problem to the left side. And the location on the scale will determine the best approach in estimating time and effort.
Another time we will talk about how to tackle open-ended problems suffering from power law disturbances.