Using scan data to choose which bins to recount

After a scan count, the table usually shows a long list of differences. Recounting every one of them would take as long as the count itself, and a recount is not free of mistakes either. A better approach is to recount a smaller number of bins and to choose them with care. The scan records already contain most of what you need to choose.

What the scan data can tell you

Some signals come from the size of the gap:

• the gap in pieces against the books
• the gap value, if you have the cost
• the gap as a share of the book quantity

Others come from how the scanning went, and these are visible in the records:

• a bin scanned more than once, or by two different devices
• a long pause inside a bin, or dozens of scans within a few seconds, which suggests a repeated or bulk entry
• a bin where the books show stock but nothing was scanned, or the reverse
• an item scanned in a bin that is not the one in the books
• a scanned code that is not in the item master

None of these signals proves an error on its own. Together they tell you where an error is most likely.

A simple scoring

Give each bin a score for the signals it shows. A starting scale:

• gap value among the highest 10 per cent of gaps: 3 points
• gap of 10 pieces or more, or 10 per cent of book quantity: 2 points
• bin scanned twice or by two devices: 2 points
• books show stock, nothing scanned: 3 points
• scanned code not in the master: 1 point

Recount the bins scoring 4 or more, highest first, and adjust the scale to your own range after the first count.

Suppose bin A-03-12-2 holds B-0412 with book 310 and counted 296. The gap is -14 pieces, so it is high in value (3) and above 10 pieces (2), and two devices scanned it (2). That is 7 points, and the bin goes to the top of the list. Bin C-01-04-1 holds A-1042 with a gap of -2 and no anomalies. It scores 0 and can wait. Bin B-02-07-3, where the books say 36 and nothing was scanned, scores 3 + 2 = 5 and is recounted. If the team can recount 25 bins in the afternoon, take the 25 with the highest scores.

Recount so that the result means something

A recount only helps if it is independent:

1. The recounter does not see the first count or the book figure.
2. It is a different person from the first counter.
3. The recount is saved as its own scan record, marked "recount", with time and operator.
4. The decision rule is written down beforehand.

A workable rule: if the recount equals the first count, the count is confirmed. If it equals the book figure, the first count was wrong. If it matches neither, a supervisor counts a third time.

Add a small random sample

Scoring finds the problems you can see. To catch the ones you cannot, add a small random sample of low-scoring bins, for example one in twenty. If the sample turns up errors, the first count is less reliable than the scoring assumed, and the threshold should come down.

Learn from the outcome

After the recounts, note for each bin whether the number changed and why: miscount, item in the wrong bin, double scan or a real stock difference. If most changes come from one cause, fix that at the source. If many items turn out to sit in neighbouring bins, the labels or the bin layout need attention, not another recount. Keep the table of outcomes; at the next count it tells you which signals were worth their points and which were noise.

Why this saves effort

With a score you can explain to the owner why 25 bins were recounted and 200 were not. You use the team's afternoon on the places where an error is most likely and most costly, and you leave a record of the reasoning. That record also makes it easier to agree on the adjustments afterwards, because each remaining difference has either been confirmed or has been looked at for a reason.

InventoryHub's scan records carry the bin, time and operator, so anomalies can be seen. Bins that do not tally are recounted on the spot, and the biggest gaps get a second count afterwards.