How does AD work increase with size of department?

Let us take a simple model

  • one project
  • one set of legacy systems running in operation

Let us call the amount of project work that the project has to do for internal purposes W. For argument’s sake let us assume that the project team only has to agree with the people outside the project team on

  • A – user requirements
  • B – data definitions and meaning compatible with legacy
  • C – infrastructure s/w levels compatible with production environment (e.g. operating system version, database management software, network protocols).

Now let us assume there are two parallel projects working on the same legacy systems. Each project has the same three things to agree with the outside world, but also to some extent it will need to agree these aspects with the other project making an additional set of three agreements. With three parallel projects this would be three lots of three external agreements plus three times three times three inter-project communications. With four projects there would be four lots of external communication and three to the power four lots of external communication. Let us call the amount of work in agreeing a subject with another project team K and the amount of work agreeing a subject with the outside world W. We can thus end up with a formula that shows that the amount of work for a set of n projects on this common infrastructure is:

  • Total work needed for n projects = W*n + (K*3)n

This would lead to an AD department productivity curve like this.


In practice big AD departments create lots of “facilitation” departments to help inter project communications such as:

  • Technical Design Authority
  • Strategy and Architecture Departments
  • Data Management Group
  • Software Release Management Function
  • AD Methods and Tools Functions
  • AD Programme Office and Control
  • Production Acceptance and OAT
  • Information Security Department

For each area requiring inter project agreement, e.g. Technical Design, each project now only has to talk to one “facilitation” department as opposed to all the other projects. This reduces the exponential nature of the formula above in our model however, the number of facilitation departments required tends to increase with the number of parallel projects, so the formula is more likely to be:

  • Total AD work needed for n projects = n*W + Kf(n)

Where f(n) is some function that is always increasing but less than n. This gives a yield curve of AD investment like this.


We believe big bank AD departments are on the flat part of the curve, where extra AD projects investment produces little or no extra AD output.

Why do projects have to talk to each other?

There are two route causes:

  • To avoid expensive to fix problems later in the project life cycle. If two projects are both going to use a new web server component they had better agree the software level early on or one of the projects is unlikely to go live without re-work and re-test.
  • To avoid long term cost of ownership issues. If in the previous point, both projects had gone live on their different levels of web server components, the IT ops function would have to support both levels for many years (e.g., duplicate s/w licences, skills, test environments, etc.)