”All models are wrong, but some are useful.” This insight from George Box has become a cornerstone of software modeling. While exploration and experimentation are essential for shaping models, it is often challenging to know when a model is wrong or, worse, when an existing model has become obsolete and is no longer as useful.

As a result, unsuitable models may be still used for a long time, leading to different issues not always identified as forms of design debt. These include the spork effect (intrinsic coupling), model fragmentation (extrinsic coupling), and model sclerosis, which ultimately contribute to entropy or the "Big Ball of Mud" phenomenon. These problems arise from undetected and unresolved model tensions.

In this talk, we will explore what model tensions are, the complications they cause and why they occur. We will also present a set of heuristics to identify and address different kind of model tensions.