Statistical Management In Film Production

versus Guessing
By Greg Passmore of PassmoreLab - The Art of Science

As many of you know, I am in the business of making films. The process of film production is largely run by creatives and lacks the management discipline (and training) found in more conventional factory operations. It is a frequent error to believe that filmmaking is a "work to quality." Instead, almost all commercial film production today is an endeavor of "work to cost." In the case of studio films, there may be the added pressure of "work to schedule." Since budgets rarely flex, schedules and quality have to become plastics. We have three points of control; quality, cost and schedule. Change any one of them and the other two move, like it or not. All of this can create a toxic environment of stress, where there exist ambiguous quality specifications, a delivery date promised by sales and not production, and the entire process is operating on a fixed budget.

This yields the classic management solution of making people work longer hours for the same pay with the expectation of level quality. This behavior, which I call "management by panic", is ubiquitous in visual effects and film production. Its energy is the aggressive drive toward deadlines made by others, quality judged by others and a pressure cooker business model of fixed costs. One visible example of this management from time/budget pressure is the recent death of Sarah Jones and the subsequent imprisonment of the film's director. 

This industry wide environment, where there is a drive to shove quality, schedule and quality all at once, leads to a lot of schedule guessing. Schedule guessing is not evil in itself, but leads to failed projects, unhappy workers, underbid projects and an environment of stress, where management by panic leads to poor judgment. 

There is an alternative. 

In visual effects, we have the ability to add code into our tools. This is called "instrumenting the software" and yields hard cold productivity data to allow accurate predictions of vfx delivery schedules. This is not a common current industry practice. However, by having unbiased data about worker productivity, shot complexity, probability of reworks, and the time trend analysis that comes with this, we can create an environment of rational and sane process control. 

If you are a manager that bullies your workers, or a worker who feels they are pushed to "just give me the delivery date!", the correct answer is in the form of probability and statistics. Brush off your old textbook and look at polynomial regressions, scatter plots and infographics. Use this to empower your project to yield beautiful precision and boring predicability. For example, when you install that new whizzy version of software, see if a productivity bump actually occurs and if so, did the savings outweigh the cost.

I feel it is time to bring visual effects and film production out of the garage and into the adult world of statistical factory management. To do anything else is unreliable and management by panic and potentially toxic to those you work with.

Management without data, is not managing at all.