When the count and the records disagree: reconciling site stock

Every stock record drifts. A use goes unrecorded, a delivery is entered twice, a hose fitting weeps, a box of tubes walks off site. Drift itself is normal. What matters is that you notice it soon, find out why, and correct the record in a way you can explain later.

Two figures, kept apart

The method that works is to keep two figures for each container and never let one overwrite the other:

• Recorded: the opening quantity plus every delivery, issue, dispense, transfer and return someone entered.
• Measured: what a physical count, a gauge, a scale or a sensor says is there.

If you simply type the count over the recorded figure, the difference disappears and with it the only evidence that something went wrong. Keep both, and the gap becomes something you can act on.

Not every difference matters

Gauges and sensors read with some error, and counts of part-used containers are estimates. A small tolerance stops every rounding difference from becoming a task. A common choice is 2% of the recorded figure: a bigger gap is reviewed, a smaller one is not. Fuel and high-value items may deserve a tighter band.

It also helps to record how each figure was arrived at: verified by weighing or counting sealed units, measured by a gauge or sensor, calculated, estimated, or unknown. A variance against an estimate deserves less alarm than one against a verified count.

Review the variance, not the people

A variance is a question: why does the measured figure differ from the records? The answers are usually ordinary:

• a use nobody recorded, often on a busy day;
• a wrong unit, such as gallons entered as quarts;
• a leak, spill or evaporation;
• a delivery that was short or entered twice;
• a sensor or gauge that needs recalibrating;
• theft, which is the least common answer and should be the last one reached.

The review should record what was found, not who is to blame. If people fear the count, they stop counting honestly.

Correct the record, keep the history

Once someone has looked, there are three honest outcomes:

1. Accept the measured figure. The record is corrected with an explicit adjustment entry that says who accepted it, when and why. Nothing earlier is edited or deleted.
2. Explain it. For example, the count was taken mid-delivery and a later count agrees with the records.
3. Dismiss it. The measurement was wrong: a misread gauge or a faulty sensor reading.

Whatever system you use, avoid one where a correction overwrites the old figure. An adjustment entry with a reason keeps the history you will need when someone asks what happened.

Find drift sooner

The longer a gap goes unnoticed, the harder it is to explain. Ways to shorten that:

• Count on a rhythm. A weekly count of the chemical cabinet or a monthly count of the main store, at a fixed time.
• Count the risky items more often. Fuel, high-value products and anything that walks.
• Put a sensor on the big tanks. An hourly level reading turns a leak found at month end into one found the same afternoon. See Tank level sensors on a construction site.
• Record at the point of use. A scan at the container beats a list at the end of the day. See QR and NFC labels for site stock.

Use what the variances tell you

Look back over a month of variances. If most are unrecorded fuel draws on Mondays, the problem is the Monday routine, not the fuel. If the same sensor keeps causing dismissed variances, recalibrate or replace it. Patterns in the variances usually point to one process fix that removes most of them.