Cycle counts beat the annual stocktake

The annual stocktake is a ritual: shut the operation, count everything, write off the difference, reopen. It produces one number — total shrinkage — and no information about how any of it happened. For an operation already running an inventory system, that is a strange thing to close the doors for.

Why the timing is the problem

An error found eleven months after it occurred cannot be investigated. The delivery note is filed, the staff have rotated, the CCTV is overwritten. All you can do is adjust the system to match the shelf and move on, which is why the same error recurs.

Errors found within a week are traceable. There is a receipt to check, a person to ask, a picking run to replay.

Count continuously, weighted by risk

Cycle counting counts a small subset every day so everything is counted several times a year without stopping work. Which subset is the interesting part — not round-robin, but weighted:

• High value first. The cost of being wrong is highest.
• High movement next. More transactions, more chances to be wrong.
• Recently adjusted. An SKU that needed a correction last month is the most likely to need one again.

count_tasks
  sku, location_id
  due_on
  priority_score      -- derived from value, velocity, error history
  assigned_to
  status

A dozen SKUs a day covers a mid-sized warehouse comfortably. The counter enters them on the same handheld everything else runs on, so the scan path has to be fast or the count gets written on paper and typed in that evening, which defeats the point.

The count writes a movement, not an override

When the count disagrees, do not set the quantity. Write an adjustment movement for the difference, referencing the count sheet:

delta: -3   reason: adjustment   reference: count_4471

The history stays intact, the correction is attributable, and the error-rate-per-SKU metric that drives the next round of priorities comes free.

The metric that matters is not shrinkage

It is count accuracy — the proportion of counts where the system already agreed with the shelf. Shrinkage tells you how much you lost. Accuracy tells you whether the system can be trusted, which is what determines whether anyone bothers with the spreadsheet they keep on the side. It also decides whether the number is safe to publish on a storefront, where the catalogue brings data problems of its own.

Track it per location and per counter. It moves quickly once people can see it, and unlike shrinkage it is actionable in the week you measure it.