, Retail
Two people spend a Sunday counting a hardware shop. They finish at half past four with a stack of sheets, and on Monday morning the bookkeeper enters the numbers. The system says the shop should be holding £96,400 of stock. The count says £88,900.
Nobody has time to find £7,500. The stock figure in the system is set to £88,900, the difference goes to a cost of sales adjustment, and everyone agrees that shrinkage is a problem. The following year the same exercise produces a gap of £8,100 and the same conversation.
The count was accurate. It was also useless, because nothing was learned from it.
A full count is expensive and it is not the only option
The traditional stocktake closes the shop, counts everything, and reconciles once a year. It exists because the year end accounts need a stock figure and because auditors, where you have them, want to attend a count.
It has two costs. The obvious one is a day of trading and a day of wages, often at premium rates. The less obvious one is that a single annual count gives you one data point a year. If it shows a problem, the problem could have started in any of the previous fifty-two weeks, and the trail is cold.
Cycle counting is the alternative and it is better for almost everything except satisfying an auditor in one sitting. You count a subset of lines every week, on a rota, so that everything is counted at least once a year and your important lines are counted far more often. The usual approach is ABC: your A lines, the small proportion of your range that produces most of your value or moves fastest, get counted monthly or weekly. B lines quarterly. C lines annually.
The benefit is not the counting. It is that a variance found on a line counted four weeks ago has four weeks of history behind it rather than a year, and four weeks of history is short enough that somebody remembers what happened.
Most retailers end up doing both: rolling cycle counts through the year for control, and a full count at or near the year end for the accounts. If you do that, cut-off becomes the thing to get right.
Cut-off is where accurate counts turn into wrong numbers
Cut-off is the boundary between what is in and what is out at the moment of the count, and it produces more errors than miscounting does.
Goods received but not yet invoiced are physically in your stockroom and financially nowhere, so they inflate your stock without a matching creditor. Goods invoiced but not yet delivered are the reverse. Stock sitting in the delivery van at five o'clock on Sunday is either counted or it is not, and both answers are fine as long as everybody uses the same one. Customer orders paid for and set aside are already sold and should not be in your stock. Goods out on sale or return, consignment stock, items with a supplier for repair and stock held at a second site are all easy to double count or miss entirely.
The fix is procedural and boring. Stop goods movement during the count. Mark a clear cut-off point in your receiving and despatch records and write the last document number down. Segregate anything that should not be counted and label it. Count areas rather than categories, so that nothing is counted twice and nothing is skipped, and tick off a floor plan as you go.
Value it correctly afterwards. FRS 102 requires inventory at the lower of cost and estimated selling price less costs to complete and sell, so damaged, obsolete and heavily discounted stock is not worth what you paid for it. A count that records 40 units of something you will never sell above half price and values them at cost is overstating your balance sheet and your profit.
Count to a system, or you are just making a list
There is a large difference between a count that produces a number and a count that produces a comparison.
Counting to a system means your point of sale or stock package holds an expected quantity for every line, built from opening stock plus receipts minus sales minus recorded wastage and returns. The count then tests that expectation. The output is a variance report, line by line, in units and in value.
Counting without a system produces a total, and a total can only tell you about your gross margin in aggregate. It cannot tell you that eleven of one line have gone missing while everything around them is correct, which is the information that actually leads anywhere.
Two practical points make variance reports usable. First, count blind. If the counter can see the expected figure, some counters will write it down, particularly at four in the afternoon. Software that hides the expectation until after entry removes an entire class of error. Second, have a second person recount any line with a material variance before it goes on the report, because a large proportion of apparent variances are counting mistakes and investigating those wastes the time you should be spending on the real ones.
Barcode scanning into a handheld is worth the money in any shop with more than a few hundred lines. Handwritten sheets introduce transcription errors, and transcription errors look exactly like shrinkage.
The variance is the product, and the temptation is to bury it
Once you have a clean variance report, the work begins, and this is the step that gets skipped.
Sort by value, not by unit count. Forty missing screws matters less than two missing cordless drills. Take the top twenty by value and find out what happened to each one. The causes are a short list and they are distinguishable if you look.
- Process failures: goods received and never booked in, supplier short deliveries never claimed, deliveries booked in twice, wastage and breakages thrown away without being recorded, staff sales and refunds put through incorrectly, transfers between sites recorded at one end only.
- Till and data failures: two similar lines sharing a barcode, so one always runs positive and the other always negative, a multipack scanned as a single, a price override that also changed the item, or a line with no barcode that staff ring through as something else.
Actual theft, internal and external, is the remaining category and it is usually smaller than operators assume before they do this work. It matters, but chasing it while your goods-in process is broken fixes nothing.
Write the conclusion down against each line. Next count, the same line either behaves or it does not, and you have a record either way.
The habit to guard against is the silent adjustment. When a count disagrees with the system, adjusting the system to match the count is correct as an accounting entry and dangerous as a management practice, because the adjustment removes the evidence. If you make the adjustment without recording the reason, you have destroyed the only trail you had. Post the adjustment, but post it with a reason code, and report the total of unexplained adjustments every month. If that number is growing, your control is deteriorating no matter how healthy the counts look.
One more discipline: never let the person responsible for an area be the only person who counts it. That is not an accusation, it is the same reason you would not let one person both order and pay.
Where this shows up in the game
Shops are part of Business Baron's full game, and e-commerce is free to run. Gross margin is calculated from actual stock movements rather than list prices, so waste and theft show up where they belong. The players who look into a variance find causes they can fix, and the ones who do not pay for it every year.