The existing property and neighbour module stops at a boundary that must now be crossed. It treats what one damages on someone else's land as one case among others. Yet on an urban operation a single family of damage dominates all the others combined, and it comes from no impact, no fire, no falling object: it is ground movement. An excavation being emptied, a water table drawn down, a pile driven, a tunnel passing underneath displace ground over tens of metres, and that displacement carries into the foundations of whatever stands around.
This mechanism runs straight into the way policies describe the event. Settlement is neither sudden nor accidental: it is slow, continuous, expected, and often calculated in advance by the designers. It has no date. Wording that presupposes a sudden and unforeseen occurrence is therefore in tension with the most frequent phenomenon of the urban site, and that is why this risk is handled by specific clauses rather than by the policy's general regime.
Where the liability comes from must then be said, because that is where much reasoning goes wrong. In French law, abnormal neighbourhood disturbance imposes liability with no need to show fault: it is enough that the disturbance exceeds what a neighbour must bear. A project owner that did everything right, followed good practice, respected the thresholds and kept its undertakings, remains answerable for the abnormal harm it causes. Having worked well is a defence to an allegation of negligence; it is not a defence against the neighbour, and the two arguments run before different judges on different proofs.
The document that decides everything predates the site, and it is the record of the condition of existing property before works. It fixes the starting line. Without it, every crack at the neighbour's is presumed new and the debate becomes unverifiable, which always favours the party claiming. With it, the discussion moves from the ground of assertion to that of counting: what is in the record was there, what is not in it appeared. That shift is the only thing that makes a ground movement file defensible, and it can only be obtained beforehand, never after.
Then comes the point most operations miss, and it turns on the quantity being measured. Site monitoring watches absolute settlement, with alert and stop thresholds expressed in millimetres, because that is what an instrument gives directly. But what cracks masonry is not absolute settlement: a building descending two centimetres as one block does not crack. What cracks is distortion, that is the difference in settlement measured against the distance over which it occurs. Works can therefore stay green for the whole of the site period on what is measured, and be well beyond what damages them.
That distinction has an immediate practical consequence for where measuring points sit. Distortion shows itself over the distance where settlement changes, and that distance is sometimes very short: the boundary between two foundation types, a re-entrant in a façade, the edge of an old cellar, the place where fill abuts a rock bench. Targets set every ten metres along a regular façade will never see a step expressing itself over three metres. The monitoring grid therefore decides what one is able to observe, and it is chosen on knowledge of the neighbouring building, not of the excavation.
That leaves apportionment, which is the real difficulty of neighbour files. The building opposite was rarely chosen by the project: it is old, it has moved before, it carries earlier defects that the record captured, and the site's movement adds to a history that began earlier. The cost of repair therefore depends as much on the neighbour's initial condition as on the size of the movement caused, and that initial condition is neither controlled nor negotiable. That is what makes an urban operation's exposure largely independent of the quality of its execution.
An urban excavation runs alongside an 1890 masonry building. The pre-works condition record, drawn up fourteen months earlier, notes 23 cracks. Monitoring carries absolute settlement thresholds of 8 mm for alert and 15 mm for stop, on targets spaced ten metres apart. At the end of earthworks the maximum settlement recorded is 11 mm: the alert was crossed, the stop never, and the site was not halted. The neighbour claims €640,000. A post-works record counts 61 cracks. The building rests partly on a rock bench and partly on old fill, the boundary between them passing 3.20 m from a façade re-entrant, where measured settlement is 11 mm on one side and 1 mm on the other. Does the project owner have a defence?
Distortion must be computed before the thresholds are argued, and it is the calculation the monitoring did not produce. Over the 3.20 m separating fill from rock, the settlement difference is 10 mm, giving a distortion of 10 in 3,200, that is 1/320. Masonry commonly cracks well before that value, the order of magnitude usually cited sitting around 1/500. At that precise place the building was therefore well beyond what damages it, throughout the time the dashboard showed 11 mm of settlement against a 15 mm stop threshold. Both statements are true together, and that is the whole subject: the thresholds bore on a quantity that does not crack. Targets ten metres apart could not, in any case, see a step expressing itself over 3.20 m, so the arrangement was blind at exactly the place it needed to look. On liability, respecting the thresholds is no defence against the neighbour, since abnormal disturbance imposes liability without fault: working well protects against an allegation of negligence, not against a claim founded on the abnormality of the disturbance. What remains to the project owner is apportionment, and it owes that entirely to the prior record. Of the 38 new cracks, those outside the zone of high distortion sit poorly with being charged to the site, and the 23 earlier ones are out of the argument because they are written down. The non-obvious point is that the decisive document was drawn up fourteen months before the first bucket of soil, and that the costly arrangement put in place during the works contributed nothing.
- 01On an urban operation ground movement dominates all other families of neighbour damage combined, and it comes from no impact and no fire.
- 02Settlement is neither sudden nor accidental: it is slow, expected and often calculated in advance, which puts it in tension with any wording presupposing a sudden and unforeseen occurrence.
- 03Abnormal neighbourhood disturbance imposes liability without fault: having respected good practice and the thresholds is a defence to negligence, not against the neighbour.
- 04What cracks masonry is not absolute settlement but distortion, the difference measured against the distance: works can stay green on what is measured and be well beyond what damages them.
- 05The record of existing conditions before works is the only document that makes apportionment possible, and it can only be obtained beforehand: without it every crack is presumed new.