Eric Evans — Tackling Complexity in the Heart of Software

Be careful of what you mean by “realistic”.

The Mercator projection of the earth to produce a map that sailors can use. You feed the model points of latitudes and longitudes, then you can measure angles between any points which will correspond to the actual angle you need to sail in. Note that the goal was never “to produce the most accurate flat projection of a round globe”. Everyone knew that that was no possible. The goal was a device that can be used to measure angles.

When we use the word “realistic” to describe models we must be specific: realistic for what? The Mercator map is realistic in the angles between locations, but is not realistic for calculating areas. When we say realistic we trick ourselves by thinking of some other metric the thing was never intended to solve. So how “realistic” a model is, is how well it can transform data into the things that matter to us.

Models are teleological. They are for something. Keeping this in mind helps you in two ways.

1. It helps you focus when building. One easily falls into the trap of trying to account for everything. Focus on what you are trying to do.
2. It helps you avoid mistakes. Mistakes occur when we forget the purpose of a given model and start substituting it indiscriminately for the real thing. e.g using a Mercator map to calculate area.