Whether Model Performance keeps up with your model depends entirely on how the model got into Pirros.
Linked from Forma
Model Performance reflects the current state of the model. Someone syncs to central, you refresh, and the numbers move.
In practice this means you can watch the effect of work as it happens — someone cleans up a category of warnings in the morning and you can see it that afternoon. It also means the number is trustworthy without anyone having to remember to re-upload anything.
Uploaded from your machine
An uploaded model gives you a snapshot, frozen at the moment you uploaded it. It does not change as people work, no matter how much the model changes. To get current numbers, upload the model again.
This is not a degraded analysis — the numbers are accurate for the file you uploaded. They are simply about that file, at that moment.
The practical consequence is that trends are much less useful. A series of snapshots taken whenever someone happened to upload does not tell you much about direction.
If you are deciding
Model health is one of the stronger arguments for linking rather than uploading. The value of this view is largely in watching it change, and that only works on a linked model.
See Add a model: Forma-linked vs. local upload for the full comparison, including the effect on scheduling.
Frequently Asked Questions
Q: Our numbers haven't changed in weeks even though the model has. Why?
A: Almost certainly an uploaded model rather than a linked one. An uploaded model is a snapshot frozen at the moment of upload and does not change until you upload again.
Q: How do I see the effect of a cleanup we just did?
A: On a linked model, sync to central and refresh — the numbers move. On an uploaded model, upload the file again.
Q: Are the numbers wrong for an uploaded model?
A: No. They are accurate for the file you uploaded. They are simply about that file at that moment, which is why trends over a series of ad-hoc uploads don't tell you much.
Q: Is this a reason to link rather than upload?
A: It is one of the stronger ones. Most of the value in model health is watching it change over the life of a project, and that only works on a linked model.