The Eiffel Tower is usually discussed as a design achievement. The more useful story for anyone planning an erection sequence is what happened at 57 metres, in March 1888, when four separately erected inclined legs had to meet a horizontal platform at the same elevation, at the same angle, at the same time.
The convergence problem
Four inclined piers, each rising independently on its own foundation, each carrying its own creeper crane, each subject to its own accumulation of fabrication and erection tolerance. At the first platform they stop being four structures and become one. Every error accumulated on the way up arrives at that junction simultaneously.
Nothing about that is unusual in principle. What distinguishes the Eiffel Tower is that the response to it was built into the erection method before the first piece of iron went up.
The adjustment system was not a remedial measure. It was part of the plan.
Two systems, working in opposite directions
Sand under compression behaves as a very stiff but adjustable bearing: it distributes load, but its height can be changed by releasing material through a controlled outlet. Drain a box evenly and the leg settles. Drain one side only and the bearing surface becomes asymmetric — the leg tilts. Repeated at intervals as the leg climbed, that gave the erection team the ability to steer the trajectory of an inclined column, making corrections invisible at the base but decisive by the time the leg reached 57 metres.
Jacks fitted to the shoes at the base of each leg worked in parallel with the sand — described in published accounts as capable of exerting a force in the order of 800 tonnes, operated by hand pump. They provided immediate positional correction where sand drainage was too slow or too coarse. Once the platform girders were riveted and the geometry fixed, the sand was progressively released and the legs settled onto permanent bearings on the masonry piers.
Accounts differ slightly on how the two systems were used together. Some describe the legs as intentionally erected steeper than required, with the sandboxes then letting each leg settle into position. Others describe the jacks raising the piers to level the junction. Both movements were available, and that is the substantive point: the erection method retained the ability to move a leg in either direction after it was standing.
Geometry control system — erection state and base detail
Tap to enlarge
Diagram: Original — EE&HL Network 2026.
What the tolerance figure actually means
Published accounts describe the first-floor girders being aligned to an accuracy of about one millimetre. That number is often quoted as evidence of nineteenth-century fabrication precision. It is better read as evidence of something else.
One account of the project records the error at the leg junction, before adjustment, at several centimetres. Whether that specific figure is exact is difficult to establish from secondary sources — but the existence of the adjustment system tells its own story. Nobody installs 800-tonne jacks and a sandbox system on the expectation that the pieces will simply fit.
Fabrication precision was real: the official tower history states the components were traced out to an accuracy of a tenth of a millimetre in the Levallois-Perret workshop. But drawing-office precision and site geometry are different problems. Four inclined piers rising on separate foundations accumulate error that no workshop tolerance can pre-empt. The millimetre at the platform was achieved on site, by a temporary works system designed to absorb the difference between what was fabricated and where it actually arrived.
Why the tolerance was not negotiable
The tower was assembled from puddled iron shaped and drilled in the Eiffel workshops at Levallois-Perret and delivered pre-drilled and pre-matched. Every hole sat in a fixed position. Every connection length was fixed. Welding would not become practical for structural steelwork for another three decades, and there was no capacity to cut or re-drill on site without compromising the connection.
That makes the first-platform tolerance a fabrication constraint rather than a quality preference. The prefabricated girders spanning between the four legs either reached their connections or they did not. Force them and bending loads enter joints designed for axial and shear forces alone. There was no field remedy.
Which is why the control system had to be specified before fabrication began, not improvised once a problem appeared on site. The geometry of the temporary condition — four freestanding inclined legs, each on an adjustable base, each climbing independently with no lateral restraint from a structure that did not yet exist — determined whether the permanent connections would be assemblable at all. Eiffel and his chief engineer Maurice Koechlin designed for that condition in advance.
The commercial reading
There is a pattern worth naming here, because it recurs on projects a century later.
Adjustment capability is cheap to install before erection and expensive to retrofit during it. The sandboxes and jacks were a known cost, planned, budgeted and scheduled. Discovering at 57 metres that four converging legs do not meet, with no means of moving them, is a different category of problem entirely — one measured in programme rather than in equipment hire.
The tower was assembled in twenty-one months. The critical junction at the first platform was completed by the end of March 1888. Both of those figures depend on the fact that when the legs did not line up, somebody could do something about it that afternoon.
There is a question underneath this that every project answers early, usually without noticing: is geometry control an erection problem or a design problem? Answered as an erection problem, the site team is expected to manage tolerance with whatever means are to hand. Answered as a design problem, the control mechanism is specified before components are made, and the erection team inherits a system rather than improvising one. That answer is very difficult to change once fabrication has begun.
The temporary works came down. They usually do, and they usually go unrecorded. The structure that remains is the one that gets photographed.
Frequently Asked Questions
Two adjustment systems worked together. Sand-filled boxes at the leg bases acted as stiff but adjustable bearings: draining sand evenly let a leg settle, while draining one side made the bearing asymmetric and tilted the leg, allowing its trajectory to be steered as it climbed. Hydraulic jacks at the shoes provided more immediate correction where sand drainage was too slow or too coarse. Once the platform girders were riveted, the sand was progressively released and the legs settled onto permanent bearings.
Accounts differ. Some state the legs were intentionally erected at a slightly steeper angle than required, with sandboxes then allowing each leg to settle into position. Others describe the hydraulic jacks raising the piers to level the first-platform junction. Both movements were available to the erection team, and that is the substantive point: the method retained the ability to move a leg in either direction after it was standing.
Published accounts of the project describe the first-floor girders being aligned to an accuracy of about one millimetre using the sandboxes and hydraulic jacks. The critical junction of the four legs at the first platform, at 57 metres, was completed by the end of March 1888.
It shows adjustment being designed into the temporary works before it was needed, rather than improvised once a problem appeared. The sandboxes and jacks were not a remedial measure; they were part of the erection method from the outset, and they were removed once their purpose was served.
Sources: Official Eiffel Tower history, Société d'Exploitation de la tour Eiffel (toureiffel.paris); PBS NOVA, Building the Eiffel Tower (2024); archival photographic records of the 1887–1889 erection; Wikipedia. Figures are reported as published in those sources. Where accounts of the adjustment sequence differ, the difference is stated in the text rather than resolved.