Physicists have long alluded to demons in thought experiments. They did not really mean the scary demons like above or commonly held in people’s minds. Instead, their demon was an all-knowing entity, with no limits on its computational powers or memory size. Laplace’s Demon could predict everything about the future, or the past given the total knowledge of the universe at any given moment. Maxwell’s Demon could defy the Second Law of Thermodynamics.

Of course, there are no demons. In theory, it is possible to calculate the path of every molecule in the air in the room you are sitting. But no person, no computer and no instrument has the power to measure the state of every molecule and then calculate their paths into the future. To get around this limit on our intellect, Boltzmann and Maxwell helped birth statistical mechanics. They showed that while it was impossible to know everything, with the aid of statistics and probability we could predict what matters: pressure, temperature, volume and entropy. These values can be inferred with only assumptions about the probability distributions of the gas’s particles state rather than their exact positions.

We need to try and find the equivalent to statistical mechanics in managing programs. Trying to keep track of every molecule and predict its path is a lost cause, as is trying to identify every task in a program and keep them on their predicted path. A schedule that is too detailed cannot be understood or used by anyone (except by a Demon). Useful schedules are ones that can tolerate unexpected events without breaking or becoming obsolete. Useful schedules provide cues to people working the program about what the most important tasks are and what where they should be spending their time.

We should be much, much more skeptical of using detailed schedules that are demanded of program managers to forecast program timeline and cost. Over and over again this practice has led to large misses. Bewilderingly we expect the next time to be different. Companies or industries should do a much better job of creating databases of historical project lengths and costs for large programs and making estimates based on these. These databases should be how we predict what matters: duration and cost given scope. If companies propose programs that are outside the distributions of historical performance, it should be accompanied by very specific reasons of why this time will be different. Unleashed from the prime responsibility of predicting cost and schedule, program plans can be made mutable, free to change with unexpected circumstances and guide the work to be done as it is, not as it was hoped to be.