Operations
When your stock and your books disagree
Most shops find the gap at year end, treat it as shrinkage, and move on. It is usually not theft, and finding that out is when margins start to make sense.
Ask a shop owner what their closing stock is worth and you will usually get two answers: the figure the accountant used, and the figure the owner believes. The difference is rarely small, and it is almost never investigated, because investigating it is tedious and the year is already closed.
The gap is not usually theft
Shrinkage is the first explanation people reach for, and it is the least common one in the shops we have looked at. What actually happens is more boring. Goods arrive and are put on the shelf before anyone records them. A customer returns something and it goes back into stock without a note. A damaged item is written off in someone's head. A loose-weight product is sold in quantities that never quite reconcile with how it was purchased.
None of these is dishonest. All of them widen the gap.
Why year-end is the worst time to find out
By the time an annual count happens, you are looking at twelve months of accumulated difference with no way to attribute any of it. You cannot tell whether the problem was in March or in October, whether it was one product line or all of them, or whether it is getting better or worse. You get a single number and no story.
A shop that counts one category a week has the same total effort spread differently, and gets something far more useful: a difference small enough to explain while people still remember the week it happened.
The measure worth watching
Not the size of the discrepancy. The time between an error happening and someone noticing it.
If that number is falling, your process is improving, whatever the rupee value says in any given month. If it is not falling, a bigger annual count will not help — you will simply be more precisely wrong once a year.
Where software helps, and where it does not
Software cannot stop someone putting goods on a shelf before recording them. What it can do is make recording them the path of least resistance: a scan at receiving that takes less effort than writing it in a book, and a running balance the person at the counter can see without asking anyone.
If entering data is harder than not entering it, people will not enter it, and no system survives that. This is the single most useful thing to test before you buy anything: watch the person who will actually use it do the thing they do fifty times a day, and count the steps.