Intermediate Precision: The Most Misunderstood Validation Parameter

Early in my career (long before I was accountable for approving validation strategies and defending them in regulatory settings), I treated intermediate precision the way many of us do during method validation, something to be demonstrated, documented, and then quickly checked off the list. Different analysts, different days, maybe a second instrument. Run the study, calculate the %RSD, move on.

With experience, that perspective changed.

Looking back, intermediate precision was never just a statistical requirement. It was always meant to answer a much bigger question: how consistently does this method perform when normal, expected sources of variability are introduced? Different analysts. Different days. Slightly different ways of executing the same procedure - exactly what happens in a real QC environment.

Repeatability tells me how a method behaves under tightly controlled conditions. One analyst, one setup, one environment. Useful, but incomplete.

Intermediate precision introduces the human element. Different hands. Different interpretations of the same procedure. Slight differences in weighing, mixing, timing, or equilibration. It is often the first point where hidden assumptions surface.

I remember reviewing an intermediate precision study for an HPLC method where system suitability passed cleanly across all runs. Retention times were stable. Resolution met criteria. Yet the %RSD crept just beyond expectations when a second analyst prepared samples on a different day.

At first glance, nothing appeared “wrong.” The instrument performed as designed. The column was qualified. The method was followed as written.

What changed was something far less obvious: how long the sample was allowed to equilibrate after preparation before injection. The method allowed flexibility. Each analyst made a reasonable choice. The chemistry noticed.

That experience stayed with me because it reinforced a pattern I have seen repeatedly. Intermediate precision rarely challenges us because the science is flawed. It challenges us because the method relies on assumptions that are never explicitly controlled.

When variability shows up, the instinct is often to look at instruments, software, or calculations. Sometimes that is appropriate. But just as often, the contributors sit upstream. Sample preparation steps that depend on technique, ambiguous instructions, or development practices that do not fully translate into routine QC execution.

This is why I no longer view intermediate precision results as pass or fail in isolation. I see them as signals.

A higher %RSD does not automatically mean a method is unusable. It tells me where the method is sensitive. It tells me where tighter procedural control, clearer instructions, or automation may be justified. And it forces an honest conversation between development and QC about what “robust” truly means.

Some of the most valuable validation discussions I have been part of did not come from perfect intermediate precision data. They came from data that made us pause, ask better questions, and strengthen the method before it ever reached routine testing.

Regulators understand this nuance more than we sometimes assume. What they look for is not artificially clean data, but a sound scientific rationale - why acceptance criteria are set where they are, how variability is understood, and how risks are managed across the method lifecycle.

Intermediate precision is not a checkbox for me anymore.

It is a mirror.

And like most mirrors, it does not always show us what we want to see - only what is actually there. The trick is resisting the urge to adjust the lighting instead of fixing the reflection.