On this page
Why age is the expensive variable
Take any recurring operational leak: a person pouring heavy, a comp code being used loosely, a menu button that rings the wrong item. The loss per occurrence is usually small. What makes it material is that it repeats every shift until somebody stops it.
So the total cost is roughly the rate of loss multiplied by how long it runs. The rate is a property of the problem and you cannot change it. The duration is a property of your process, and it is entirely within your control.
That is the whole argument, and it does not depend on any claim about how good the analysis is. A worse analysis delivered on Thursday beats a better one delivered in four weeks, on cost alone.
The second cost: the conversation degrades
There is a softer effect that operators consistently underrate.
A conversation about last night is specific and low friction. Both people remember the shift, the evidence is fresh, and it reads as coaching. The same conversation about a pattern from five weeks ago is a different event: nobody remembers the specifics, it arrives with a month of accumulated total attached, and it reads as an accusation.
The delay converts a routine correction into a confrontation. That is part of why standards slip in operations that only look at monthly numbers, and it is a cultural cost rather than a financial one.
What genuinely has to wait for the period
Not everything should be real time, and pretending otherwise produces noise.
- Anything requiring a physical count. Variance against theoretical usage needs stock on hand, and that arrives on the counting cycle by definition.
- Financial close. Prime cost, margin and the profit and loss are period concepts and should stay period concepts.
- Anything where a single shift is not evidence. One quiet Tuesday is not a trend. Reacting to it is worse than waiting.
That last one matters. The argument for speed is not that every observation deserves an immediate response; most do not. It is that the delay should be set by how long the pattern takes to become real, not by when the reporting cycle happens to run.
The practical split
The workable arrangement is two cadences rather than one.
Continuous, from the transaction stream: per person outliers against the team, void and comp patterns, product mix shifts, and anything already seen in previous weeks. None of this needs a count and all of it is available as it happens.
Periodic, from the count and the books: variance against theoretical, prime cost, margin by product, supplier price movement. These are health checks and they belong on a cycle.
Most operations run only the second and treat it as both. That is the gap, and it is expensive in proportion to how long the cycle is.
Speed on its own is not the fix
Worth being clear, because it would be easy to read this as an argument for faster dashboards. A finding delivered instantly and then ignored has cost you the same as a finding delivered late.
Speed only pays if the finding turns into work with an owner and a date. Otherwise you have built a faster way of producing things nobody acts on, which is the failure mode described in the data to action gap.