The Local Loop Principle

If management (or QA) wants to get figures out of a process, they eventually will get figures. Whether these figures have anything to do with reality is another story.

At some conference in India in 2003, people were talking about management dashboards. In my mind I got the image of a manager sitting behind an empty desk, with only a screen with the dashboard on it. Green-Yellow-Red traffic lights, some meters, and I'm sure that the manager also would like to have some knobs, to tune the process he was trying to "control".

In projects I always remind people to keep the colour of their project green at all times, because if it would become yellow, higher management would come down to poke in the project until the light would become red. After all, most managers are not better at running a project successfully than we are, are they?

Now, in order to feed the dashboard, management needs data from the project, so they ask for data. Of course nothing happens. Then they ask louder, and louder, until eventually they get data. They asked for data, they get data. What this data has to do with what actually is happening in the project is unknown.

If management starts managing with their dashboard, based on numbers of unknown relevance, we may expect problems. From Control Theory we know (at least I've learnt) that if we have a loop with high amplification (management) and a slow time-constant (management action and reaction), we may get instability. If the reaction is even based on data of unknown relevance, the instability will be even worse.

From Control Theory we also get the answer what to do to get a stable system: create a local loop, with appropriate time-constant: letting the people in the project generate data they can and will use themselves for their own benefit. If it's for their own benefit, there is a chance that this loop will be useful and sustained. If we now tap the data from this loop to feed the management dashboard, the data will probably have some meaning and the system can become stable, the large-loop time constant can be even longer, and with much less amplification.

So, I would advise, for any measurements you wish to get reliable numbers from, better help people to help themselves with a local loop and optimize that loop. If that succeeds, you can tap good numbers for management use. If this process is well in place, then your organization will benefit.

Don't try to "control" the project. Coach the project to be successful themselves. Now management will have a nicer job too.

See also the "Simple Model of Management".