Why the number in the system is wrong
Stock records do not drift because people are careless. They drift because a dozen ordinary events each move stock without moving the record.
Every business that holds stock eventually discovers that the system says forty and the shelf holds thirty-three. The usual first reaction is to suspect theft, and the usual reality is considerably duller.
Many stock problems are cheaper to prevent than to fix after the fact. For a broader workplace example of acting before problems become urgent, see this article.
Records drift because stock moves in more ways than anyone designed for.
The ordinary causes
- Received quantity assumed rather than counted. The order said twelve, ten arrived, twelve went into the system.
- Items taken for internal use — a sample, a repair, a demonstration — without a movement being recorded.
- Returns put back on the shelf and not back into the record, or recorded twice.
- Two similar items confused at picking, which creates a shortage in one line and a surplus in another.
- Damage discarded without a write-off.
- Units of measure: a case of six counted as one, or one counted as six.
Only one of those is dishonesty, and it is rarely the largest. The last one is the most common single cause in businesses that buy in cases and sell in units.
Variances that go both ways point at process errors. Variances that only ever go one way point at something else, and the distinction is worth making before anyone is accused of anything.
Every movement needs a record, including the boring ones
Most systems handle purchases and sales correctly. The gaps are in the movements nobody thought of as transactions: internal consumption, samples, damage, returns, transfers between locations, and stock used in a job rather than sold.
For a broader reference on product and location identification, GS1 standards document widely used supply-chain identification standards.
Each needs a route that takes ten seconds. Where recording a write-off requires a manager's approval and a form, damaged stock goes in the bin unrecorded, and the record drifts by exactly that amount.
Count something before you conclude anything
The size of the problem is unknown until it is measured. Counting twenty fast-moving lines takes an hour and tells you whether the record is roughly right, badly wrong in one category, or wrong everywhere.
That answer determines what to do next, and it is a different answer in almost every business.
Fix the process before the numbers
Adjusting the system to match the shelf is necessary and it is not a fix. Unless the cause is removed the same variance reappears within months, and repeated adjustment eventually trains everyone to treat stock figures as approximate.
The order that works is to count, investigate the largest variances, change the process that caused them, then adjust. Adjusting first destroys the evidence.
Accept that it will never be exact
Perfect accuracy is not a realistic target and pursuing it costs more than the errors do. What matters is knowing the size of the error, keeping it stable, and having it small enough on the lines where being wrong is expensive.