Provenance

Provenance is the recorded history of a piece of data or a result: where it came from, which steps it passed through, and which assumptions were made along the way. In a data context it means you can walk back from a number to the input it rests on.

What it means for the calculation

A calculation without provenance cannot be defended. When a customer asks why a part costs this amount, "that is what the system works out" is not an answer. Provenance makes an amount traceable to the inputs and assumptions under it, so the estimator can show which assumption drives the price and which ones they changed themselves. It also makes a difference with the post-calculation usable: you see which assumption was off, not only that the total was off.

Where this comes up

In data quality and in AI systems, where it decides whether a result can be checked. Subduxion uses the term for its own position: a calculation shows its basis, so the estimator can confirm or correct it.

Common mistakes

Confusing provenance with a log file: a list of what happened does not yet say which assumption drives the result. Also presenting a result as proven because its provenance is visible; provenance shows what something rests on, not that it is right.

Is provenance the same as an audit trail?

An audit trail records actions; provenance records the chain of input, assumption and result.

Does provenance make a result correct?

No, it makes it checkable; someone still has to judge it.

Why does this count in estimating?

Because an amount you cannot trace back is an amount you cannot defend to a customer.