Somewhere in one of our stores, someone scanned a barcode into a quantity field. The number entered was 8,388,607. At $19.99 per unit, that made a single line item worth roughly $167 million. The number was so large that the import failed immediately.
That is my favorite bug of the month because it screamed. It fell over at the exact moment it happened, costing me about 10 minutes.
Software can tell you what is in a report, but it does not automatically know what should be there.
The absurd number was the easy part. The more pressing problem was that we had recorded the wrong sales data. I knew that because of something no software had to tell me: Nobody was ever going to hand us $167 million at a register. The transaction had been canceled, voided, or refunded, but our reporting still treated it as a sale.
That is not a technical observation. It is a retail observation. And that distinction matters for every store owner who relies on software to make decisions about buying, pricing or inventory.
Knowing What ‘Normal’ Looks Like
If you have been following this series, you know that over the past year, I have been developing Retalyz, a system designed to help our chain manage inventory, ordering, margins and store performance. This summer, I began expanding it from four test stores to all 21 locations.
I expected the biggest challenge to be processing more information. Instead, the problems most likely to hurt us were the ones that looked perfectly normal on screen. The reports ran, the numbers appeared, and nothing flashed red. The results were simply incomplete or misleading.
That is where a store operator has an advantage over the software.
Software can tell you what is in a report, but it does not automatically know what should be there. It does not know that one location typically sells 40 units of an item each week, that a manager usually places orders on Tuesday, and that a negative inventory number often indicates a receiving mistake. You know those things because you know your business.
That means you do not need to be a programmer to identify many reporting problems. You need to know what healthy, normal activity looks like in your store and question anything that does not fit.
Silence Is Still a Signal
One example involved margin alerts. When we expanded Retalyz, several new stores were added correctly, but one automated check still looked only at the original four stores. The new locations received no margin warnings.
That sounds like an obvious failure, but it did not look that way. A store with no margin alerts can look like a store where every product is priced correctly. In other words, the problem looked like good news.
The simplest way to catch that kind of mistake is to verify the result after a system change. If you add a location, ask to see a real example of every report or alert the location should now receive. Do not assume that a blank report means there was nothing to report.
We encountered a similar issue when importing sales history for a newly added store. A data file stopped downloading partway through, but the system accepted the partial file as complete. The store ended up missing nine months of history. The problem was exposed not by a sophisticated test. We compared the newest sales record for that store with those of our other locations. If 20 stores show sales through last week and one stops months back, something is wrong.
Think Before You Filter
Another issue arose from trying to speed up inventory checks. Our original process reviewed more than half a million product records per store, including many discontinued items. Across 21 stores, the process was far too slow.
The obvious shortcut was to check only items with more than zero on hand. Technically, that made sense. However, from a retail standpoint, it would ignore negative quantities.
Negative inventory is not meaningless bad data. It can indicate that merchandise was sold before a shipment was entered into the point-of-sale system or that someone made a receiving error. Those negative numbers are clues. Removing them from a report can make a store look cleaner while hiding the exact problems you need to find.
Then we discovered the same issue with zeros. If a system only looks for products with inventory on hand, it cannot tell you when an item has reached zero. But zero is exactly what a buyer needs to see when checking for stockouts.
Nothing had crashed. The software did what it was asked to do. The question was too narrow.
That is a useful habit for any retailer making changes to reports or inventory settings. Ask what will disappear when you apply a filter. Before you exclude zeros, negative values, older products or certain transaction types, make sure you understand whether those records might reveal something important.
When Worse Numbers Are Better
Two of our fixes actually made our performance look worse. We found that sales revenue was recorded before discounts, which made some margins appear higher than they were. We also discovered that parked and voided transactions were counted as revenue. Correcting those issues removed $210,342 from our historical sales totals.
The reported margin and revenue both declined, but the business itself remained unchanged. The main difference was that the reports had become more transparent and accurate.
Most retailers do not need to build their own software to spot anomalies. Start with a few basic questions: Does this number make sense for my store? What should be here but may be missing? What data is being excluded? If I compare this location with a similar one, could something look unusually quiet?
Technology is excellent at answering the question you ask it. Your value as an operator is knowing when to ask another question.
A clean report and a faulty report can look identical on screen. Telling them apart starts with knowing your operation.
Zondre Watson is the general manager of technology and analytics for adult retail chain Ero-Tech. With a background in finance, chocolate and controlled chaos, he blends retail know-how with AI tools to keep 17,000 products moving smoothly.
