The half-level problem

There is a category of problem we now recognise quickly, and we call it the half level after a museum whose extensions meet between floors. The building has a true description. The true description is also unusable, because it requires the visitor to hold a model of the architecture in their head before they can take a step.

The failure mode is always the same and it always comes from good intentions. A previous team looked at the situation, understood it correctly, and represented it faithfully. The signs were accurate. Nobody could use them. Accuracy and legibility were in direct conflict and accuracy won, because it is the one you can defend in a review.

Organisations do this to themselves constantly outside of buildings. Navigation that mirrors the org chart is accurate. Product families named after manufacturing routes are accurate. A fare system with zones that follow the historical operator boundaries is accurate. In each case the structure being described is real, and in each case describing it is a decision to hand the reader a problem that belongs to you.

The move is to choose the fiction that is useful and then commit to it completely, including the maintenance cost. Name zones after what is in them and accept that you will reprint eleven panels when the collection moves. Cap the fare and stop explaining boundaries. A half-committed fiction, where the honest structure keeps surfacing in URLs or on the back of a ticket, is worse than either option.

And be specific about who is paying. In every case above, the choice moves effort from the reader to the organisation. That is the correct direction, but it is a real transfer, and a system that pretends otherwise will be quietly abandoned by the people who inherit it.