Anatomy of 16 months
of BIM coordination

What is a clash?

BIM (Building Information Modelling) is how buildings are designed today: instead of separate drawings, each specialist builds a three-dimensional digital model of their own part, which is then combined with the others to check that everything fits together.

In a BIM project, every discipline (architecture, structure, building services, fire protection, warehouse systems) models its own part of the building independently. When the models are combined, geometries sometimes overlap: a pipe running through a beam, a sensor pushing into an acoustic panel, a rack encroaching on a fire exit.

These overlaps are called clashes. They have to be found, classified, and resolved one by one.

On a real integrated-logistics warehouse project of roughly 75,000 m², 16 months of coordination recorded 25,682 interferences, spread across 39 different combinations of disciplines.

This is their story.

How to read a clash report

A clash report is a living document, not a single snapshot. The test that looks for interferences is re-run at intervals, one pair of disciplines at a time: each report is therefore a snapshot on a given date.

Every clash present has a status (new, active, reviewed, approved) that says where it is in its processing. But the most important thing is this: a report lists only the clashes that still exist. When one is resolved it does not get a "closed" stamp: it simply disappears from the next report.

This is why a single report is not read as a to-do list, but by comparing the sequence: what appears, what persists, what disappears. That is exactly what this story does, across 91 reports and 16 months.

1. The scale

Every dot on the right is a clash. Its horizontal position is the height at which it sits in the building (from -2 metres below the floor up to 75 metres at roof level); its vertical position is the depth of overlap between the two elements.

The colour marks the status: new, active, reviewed, approved.

The cluster at height zero is the ground floor. The band at 5–8 m is the first operational level. The dots high up above 50 m are the roof sprinklers.

2. The 39 pairs

Every clash pits two different disciplines against each other. Ordered by volume, seven pairs alone account for 70% of the 25,682 total interferences.

The winner, MHE vs STR, is the "warehouse racking against structure" pair: this is where the structural design meets the logistics equipment, and it is the main coordination battleground for a building of this size and function.

The number of different files and dates next to each pair indicates how many times the test was re-run over time.

The abbreviations denote the technical disciplines. For example ARC = architecture, STR = structure, MEP = mechanical and plumbing services, ELE = electrical, FIR = fire protection, MHE = material handling equipment (racking and conveying systems). Other abbreviations correspond to models supplied by the various subcontractors and denote systems specific to the project.

3. The project's pulse

Aggregating the 91 reports (each report is an automated check by the coordination software, a "snapshot" of all the interferences present at a given moment) by month reveals the rhythm of the work. It is not uniform.

The July 2023 peak, with 8,716 clashes, is the first big coordination push, when all the models are combined seriously for the first time. From there the peaks fall and spread out.

The empty months (April 2024) correspond to periods of work internal to the individual disciplines, when no interdisciplinary tests are run.

4. Four visible statuses, one invisible

Every interference still present in a report has one of four statuses: new, active, reviewed, approved. The chart shows how they are distributed, month by month.

Once the first wave of discovery has passed (in the early months almost everything is "new"), the "active" status dominates for most of the project: a clash report is, by its nature, the list of problems still on the table — and that is what makes it the tool for monitoring coordination. The "approved" status, by contrast, is rare and appears only in waves.

But the most important status is missing, and it cannot be seen here: resolved. As noted, when a clash is really fixed it simply no longer appears at the next check: it disappears, it does not change colour. This chart shows only the interferences still open; the resolved ones have already left the list.

How many are they? That is the question of the next chapter.

5. The 96% that disappears

Here is the invisible status from the previous chapter. Comparing each test of a pair with the next, on average 96% of the interferences found have already disappeared at the following check: resolved in the meantime. Only 4% are found again in the next report.

This is not a general law of BIM (it is what emerges from this data), but it says a lot about the process: the great majority of problems close quickly, between one review and the next (the cycle in which the team goes through the interferences and corrects them). What really matters is the 4% that persists.

6. The 38 zombies

Most clashes die quickly: they are resolved and never reappear. Some, though, persist: they come back report after report without ever closing. We have called them zombies.

Across the whole project we count 38 of them: interferences that appeared in three or more consecutive checks without ever reaching "resolved". Each line on the chart is one of these; its length says how many days it survived, its colour which two disciplines it sits between. The most stubborn, up to 296 days (almost ten months), all belong to the same pair.

Zombies are the cases where coordination got stuck: usually a quick fix to the model is not enough; it takes a decision, a meeting, a clarification between the two teams, to unblock them.

7. The speed of the process

"Speed of the process" here means something specific: how often a pair of disciplines re-checks itself and closes its own conflicts, the cadence at which it iterates.

For each pair we measured the average number of days between one check and the next. Each bar is a pair: the shorter it is, the more tightly that pair works. The colour says how much it actually closes: green = a high percentage of interferences resolved, blue = low.

The message is simple: the pairs that re-check often close more. The fastest pairs turn around in a few weeks; the slowest let as much as eight months pass between checks — and that is where problems pile up.