Every stock figure is a claim: the system says 148, and somewhere a shelf either agrees or does not. Inventory accuracy is how often the shelf agrees. Shrinkage is the value that has gone missing between the two.
Neither is a one-off problem. Records drift from reality a little every day, through a hundred small events nobody records. The job is to measure the drift and close the gaps it comes through.
Two measures
Inventory accuracy is the share of counted products whose recorded quantity matched what was found:
Count 200 products and find 174 exactly right: accuracy is 87%. Many businesses allow a small tolerance, such as plus or minus 2% on bulk items, and count a product as correct inside it. Decide the tolerance in advance and apply it to every count.
Shrinkage is the value lost, expressed as a share of what the records said you held:
Records say €180,000, the count finds €176,400: shrinkage is 2%. Retail typically runs between 1% and 2%. A warehouse that only supplies its own production should be well under 1%.
Accuracy counts errors in both directions, so a product with 10 too many is as wrong as one with 10 too few. Shrinkage nets them off and tells you what the errors cost. You need both: a warehouse can show low shrinkage and terrible accuracy, when surpluses on some products hide shortages on others.
Where the losses come from
Theft is the cause everyone thinks of first and it is rarely the largest. The usual list, roughly in order of size for a business that does not sell from an open shop floor:
- Unrecorded use. Samples, parts taken for a repair, consumables used internally, breakages swept up and thrown away. Each one is a stock-out nobody entered.
- Receiving errors. A delivery note signed for 500 and recorded as 500 when 480 arrived. The error enters the record on day one and is found at the next count.
- Picking and recording errors. The wrong variant picked, a quantity typed as 100 instead of 10, a transfer recorded once instead of on both sides.
- Damage and expiry. Goods that can no longer be used but are still in the record at full value.
- Theft, internal or external.
- Data errors. Duplicate products, a unit of measure confusion between boxes and pieces, an import that doubled a line.
The reason this list matters is that each cause has a different control. A locked cage does nothing about a receiving error.
Five controls
1. Every movement has a name and a reason. No stock changes without a record of who changed it, when, and why. This does not stop errors, but it makes every error traceable to a process, and traceable errors get fixed.
2. No negative stock. When a stock-out would take a location below zero, refuse it and find out why. A negative balance is a receiving or transfer error that has not been recorded yet, and allowing it buries the evidence.
3. Receive against the order, count the units. Stock enters the record only from a delivery recorded against a purchase order, with the quantity that was counted, not the quantity on the note.
4. Count small and often. A cycle count of one aisle a day finds errors while the cause is still recent. The variances it finds are the list of processes to fix.
5. Separate who can do what. The person who receives goods should not be the person who approves the order. The person who records a write-off should not be the only person who sees it. Warehouse staff who move stock do not need to see what it is worth. Separation is not about suspicion; it removes the opportunity for an honest error to go unnoticed.
Corrections and write-offs are not the same thing
A count finds 90 units where the record says 100. Two different things could be true. Either 10 units were used and never recorded, in which case the record should go down and the value with it. Or a stock-out was entered for 100 when only 90 left, in which case the record should go up. Putting the record right is not the same as deciding where the value went. Keep the two kinds of correction apart, with a reason on each, or your shrinkage figure becomes a mixture of real losses and typing errors.
How StockBold supports this
In StockBold every change to stock is a movement with a date, a reason, a location and the name of the person who recorded it, and the movement history can be filtered by product, supplier, location, reason and date. A movement that would take a location below zero is refused, even if it is backdated. Stock enters from deliveries recorded against approved purchase orders, with the quantity you enter for each line and date.
Stock counts are created per location and can be narrowed to a supplier, category or department. While counting, the Variances tab and the net variance figure show how far the record is from the shelf, and approving the count posts one movement per difference, named after the count, so every correction stays traceable. Corrections go through two different movements: an adjustment only ever adds units back, and only up to what was recorded as a stock-out for that product at that location in the year; a write-off is a stock-out with a comment, so the value comes down and the reason stays on record.
Around 40 individual permissions decide who records movements, who receives, who approves orders and who runs counts, and anyone without the permission to see values sees quantities only.
Read more about inventory management in StockBold, or the help guide on reading and correcting the movement history.