It Worked in the Model
Why performance in analysis doesn’t always translate to performance in the field
Models are useful.
They simplify complex systems into something that can be analyzed, compared, and ultimately designed around. In geotechnical engineering, that typically means representing soil behavior, loading conditions, and boundary assumptions in a way that produces a workable solution.
Under those conditions, the model often performs well.
What the model actually represents
A model is built on inputs. Soil parameters are selected, groundwater conditions are defined, and boundary conditions are applied. The output reflects how those inputs interact under assumed conditions.
Those inputs are based on data, but they also involve judgment. Soil properties are interpreted from limited testing. Variability is reduced to representative values. Conditions that change over time are often treated as static.
The result is a controlled environment.
What changes outside the model
Once construction begins, that control is gone. Soil conditions vary more than anticipated. Moisture content shifts. Installation methods introduce disturbance. Sequencing changes the stress state of the ground.
The field doesn’t operate on fixed inputs.
It operates on whatever conditions are actually present at the time of construction.
Where the disconnect shows up
When something doesn’t perform as expected, the model is often referenced as a point of comparison. The conclusion is usually that the field conditions didn’t match the assumptions.
That’s usually correct.
But it also highlights the limitation of relying too heavily on idealized inputs without accounting for how those assumptions might vary in practice.
What improves outcomes
Models are still essential. The issue isn’t their use—it’s how their results are interpreted.
Projects tend to perform better when:
inputs are evaluated for sensitivity, not just accuracy
variability is considered alongside representative values
construction methods are evaluated as part of the system, not after the fact
model results are treated as guidance rather than fixed outcomes
The goal isn’t to make the model perfect. It’s to understand where it might not hold.
Final thought
It worked in the model.
That doesn’t always mean it will work in the field.
About the Author
Nathan McNallie is a senior geotechnical consultant focused on report review, construction advisory, and identifying project risk before it becomes a field issue.
