Your stock record is probably wrong, and there is research that says so

The study

In 2008, Nicole DeHoratius and Ananth Raman published an empirical analysis of inventory record inaccuracy in *Management Science*. They examined nearly 370,000 inventory records across 37 stores of a single retailer, and found 65% of them to be inaccurate.

Two things about that setting are worth holding onto. It was one large retailer, not a survey of the struggling. And it had a computerised inventory system — this is not a finding about businesses that keep no records. It is a finding about what happens to records that are kept.

Why "inaccurate" is worse than it sounds

A record does not have to be wildly wrong to cause a problem. It has to be wrong in the direction of optimism, on a line with little stock.

If your system says four and the shelf holds two, nothing happens for most of the day. Then someone promises three to a customer. The cost of that error is not the two missing units; it is the promise, the phone call, and the customer who now knows your stock figure is decorative.

This is why overall accuracy percentages understate the damage. The errors concentrate where the stock is thin, and thin stock is exactly where a wrong number gets acted on.

What made it better, and what made it worse

The more useful half of the paper is not the headline. The authors identify factors that move the number in each direction:

• Inventory auditing practices mitigate record inaccuracy. Counting, done as a practice rather than an annual event, measurably improves the record.
• Complexity makes it worse — both the complexity of the store environment and the structure of the distribution arrangement feeding it.

And the covariates they identify explain 67.6% and 69.0% of the variance in inaccuracy across stores and across product categories respectively. That is the finding that should change how you think about it: the inaccuracy is systematic, not random noise. It has causes, the causes are identifiable, and identifiable causes can be addressed.

What this implies for a shop in Addis

Be careful about the inference, because it is easy to overreach. The study measured one US retailer. It does not tell you the accuracy rate of Ethiopian shops, and nobody has published that figure.

What it does establish is a floor for reasonable expectations. If a large retailer with dedicated systems and staff runs at 65% of records inaccurate, a business reconciling three branches by hand on a Sunday has no structural reason to do better — and several reasons to do worse. Assuming your own record is accurate because nobody has complained is not a position the evidence supports.

The practical reading

• Treat the stock figure as an estimate with an error bar, not a fact, until it has been counted recently.
• Count in slices, continuously, because the research supports auditing as the mitigation.
• Post counts line by line so you can see which items drift, not just that the total was off.
• Reduce complexity where you can. Every extra place a unit can sit is another place the record can diverge from it.

There is a separate, widely repeated claim that "the average retailer has 63–65% inventory accuracy," usually attributed to an RFID research lab. That is a different claim from this one — records inaccurate is not the same as accuracy rate — and we could not locate a publication behind it. The paper above says what it says; it is enough.