The same recipe, four percent less output than last year

A recipe has an expected yield. Individual batches vary for understandable reasons, and each variation is explained and forgotten.

Nobody sees the trend

Because yields are recorded per batch and reviewed per batch. A slow decline across a year never presents itself as an event, and by the time it is noticed the current figure has become the assumed normal.

Where drift comes from

Usually equipment condition, ingredient specification changes, or a working practice that shifted quietly. All three are addressable once identified, and none announces itself.

Revising standards honestly

Sometimes the standard is simply wrong and should be updated. What matters is that the revision is a decision made with evidence, rather than the standard being quietly ignored because everyone knows it is unachievable.