A correction curve is a small edit with large consequences. Nudge a few points and you have changed how much liquid every affected transfer physically delivers, which means you have changed results, not just a setting. That is the definition of a change worth controlling, and yet curves are routinely edited in place, in a spreadsheet cell or a dialog box, with no record of what they were before or why they moved. Treating a curve edit as a versioned event rather than a quiet adjustment is what separates a class you can defend from one you can only apologize for.
Why a curve edit is a real change
Most parameters in a class shape behavior; a correction curve determines the delivered volume directly. Edit it and a method that dispensed 50 microliters yesterday may dispense 51 today, and every downstream number, every concentration, every standard curve, moves with it. If a result generated last month is ever questioned, the first thing you need to know is which curve produced it. Without a version history, you cannot answer that, and an unanswerable question about delivered volume is exactly the kind of gap an audit exists to find.
What a versioned curve edit captures
Versioning a curve is not bureaucracy for its own sake; each field it records answers a question you will eventually be asked.
- The before and after: the previous curve and the new one, so you can see precisely what moved and by how much.
- The reason: why the edit happened, a drift in verification, a new tip lot, a temperature change, so the change is intentional rather than mysterious.
- The author and time: who made the change and when, so the record is attributable rather than anonymous.
- The evidence: the measurements the new curve was fitted and confirmed against, so the edit is justified by data and not a hunch.
- A path back: the ability to return to the prior curve, because sometimes a new curve turns out worse than the one it replaced.
That last point is the one people underrate. Tuning is not monotonic. A curve you were sure improved things can fail a later confirmation, and if the previous version is gone you are re-deriving it from memory. A history makes the rollback a click instead of an afternoon.
The habit worth building
Even without a system that enforces it, the discipline is straightforward, and it is the same discipline the change-control practices in regulated labs already assume. Never edit a curve in place without capturing what it was. Write the reason next to the change. Keep the verification run that justified it. Make sure the record says who and when. The reproducibility literature keeps arriving at the same conclusion from different directions: a shared definition of a class is not enough if you cannot say which version produced a given result, because the version is the part that actually ran.
The cost of skipping this is quiet and deferred. Nothing breaks the day you edit a curve without a record. It breaks months later, when a result is challenged, a lot is investigated, or a method is transferred, and the honest answer to which curve was in force is a shrug. By then the edit is long past reconstructing.
A correction curve edit changes delivered volume, so it changes results. Record what it was, why it moved, who moved it, and the data that justified it, or accept that you cannot say what produced last month's numbers.