The 54cm Error: How a Bridge Teaches Us About Data Interpretation
This is issue #007 of The Missing Header
Data problems don’t just happen in IT departments. Even in civil engineering, interpreting data correctly is crucial — and mistakes can cause enormous trouble.
In 2003, a bridge called Hochrheinbrücke, connecting Germany and Switzerland across the river Rhine, was built. The project was handled from both sides: one half constructed by German engineers, the other by Swiss engineers.
During construction, it became obvious that something was off: one half of the bridge was significantly lower than the other. Had they continued building, the two parts would have missed each other by 54 cm — more than 20 inches.
If you'd expect perfection from anyone, surely it would be from German and Swiss engineers, right? So how could this happen?
When building a bridge across a national border, both sides need to agree on a shared reference point for height. In engineering, this is typically given in terms of "height above sea level." But here's the catch: sea level isn't universal. It changes with time and geography, so each country uses a defined reference point.
In Germany, it's the Amsterdam Ordnance Datum.
In Switzerland, it's the Repère Pierre du Niton, a rock in Lake Geneva calibrated to a French Mediterranean tide gauge.
These “sea levels” differ by 27 cm.
The Swiss engineers, to their credit, knew about this and accounted for it in their calculations. But they made a sign error — they corrected in the wrong direction. Instead of building their half 27 cm higher, they built it 27 cm lower.
The result: a 54 cm vertical mismatch.
What this teaches us about data
The engineers actually avoided most of the common mistakes in working with data:
- They used precise definitions instead of vague terms like "above sea level."
- They were aware of the limited validity of those definitions.
- They even applied a conversion between systems.
The only thing they missed?
That errors tend to propagate.
You may remember the UTF-8 misinterpreted as Latin-1 example from last week, where café turns into café. Encoding bugs like this often double down: misinterpret the already-garbled string again, and you get cafÃ?©. Don't believe it? Try googling "cafÃ?©"…
But back to the bridge: the error had the potential to become a very expensive problem. What saved the day was not perfection, but vigilance. The engineers monitored the build closely and discovered the misalignment early enough to adjust course.
That’s good engineering — and good data practice, too.
🧮 The Missing Number
20 — The number of different sea level reference points in Europe
Thanks for reading,
Stefan
PS: You may be receiving this newsletter because you subscribed to the Tablecruncher Newsletter some time ago. I’ve rebranded it to avoid confusion with the software project. Same author, same scope — still all about solving messy data problems.