13 years in fresh: What the field has taught us, and what HeadQuarters don't see
There's a scene that repeats itself in every country we work in.
At headquarters, someone opens a dashboard. The stock numbers are there, the sales are there, the alerts are there. The plan looks solid. The indicators are green, or at least they can be explained.
400 kilometers away, in a store, a fresh department assistant manager is looking at his shelf. He sees three things the dashboard can't: a batch that won't make it through the day, a SKU showing in stock that's nowhere to be found, and a pallet that arrived late and absolutely has to be processed before 10 a.m.
Both people are right. It's the same store, but one sees it as it's described, and the other sees it as it actually is.
The System Is Never Fully Up to Date — and It's Least Accurate in Fresh
Everyone in the store knows why: a living product, handled and restocked ten times a day, has an inaccurate theoretical stock count before the morning is even over. What's less understood is exactly how big that gap is. And what it costs. A well-run store would be one where the system tells the truth.
The research says otherwise. A study published in 2026 in the Journal of Business Logistics by Yacine Rekik, Rogelio Oliva, Christoph Glock, and Aris Syntetos compared theoretical stock levels to actual stock levels across roughly 24,000 SKUs in eleven grocery stores, based on 150,000 inventory counts.
The result: 64.7% of SKUs show a discrepancy between what the system says and what's actually there. Only a third of product records are accurate.
And fresh is the worst offender. Perishable products show an average inaccuracy rate 6.4 points higher than non-perishable products. The faster a product turns, the more it's handled, the more often it's restocked, the less the system actually knows where things stand.
Work carried out for ECR Retail Loss by the same researchers points the same way, on a larger sample: 233,000 SKUs audited across seven European retailers in four countries, with 59.5% of SKUs showing a discrepancy. Among grocery and general merchandise retailers, that share climbs to 63.4%.
These numbers don't describe poorly run stores. They describe how a store normally operates.
The Cost of the Gap
The most interesting part of this research isn't the finding itself — it's what happens when you reset the count.
In the ECR study, correcting the discrepancies produced a sales increase of between 4% and 8% depending on the retailer, averaging close to 6%. For the SKUs with the worst records, the increase topped 14%.
In the Journal of Business Logistics study, an audit's effect on sales is nearly three times stronger for perishable products than for non-perishable ones. And the effect is concentrated in one specific case: SKUs the system believes are still in stock when they're actually gone. The product is missing from the shelf, no one reorders it, and the customer leaves without it.
Let's be clear about what these numbers are and aren't. They're findings from independent, third-party research on inventory accuracy — not a vendor's promise. They don't say "buy this and gain 6%." They say something more useful: the gap between the shelf as it actually is and the shelf as it's described has a cost, that cost is measurable, and it's highest in fresh.
Why Twenty Years of Data Haven't Closed the Gap
The natural conclusion is that the industry needs better systems. That's exactly what it has done, consistently, for the past twenty years. The results are real, but the gap hasn't closed. Three reasons keep coming up, and none of them has anything to do with the quality of the tools.
Pace. A major project is measured in quarters. A fresh shelf is decided in hours. Between the moment a problem is flagged and the decision that fixes it, more time often passes than the shelf life of the product in question.
Granularity. A system thinks in SKUs: vine tomatoes, 24 units in stock. A shelf doesn't work that way. Of those 24 units, six need to be sold today, and the rest will hold until Thursday. Same code, same price, same line in the system — and two different decisions to make. That difference is invisible from a screen.
Adoption. This is the hardest point, and the least measured one. A tool the teams don't use produces no reliable data, therefore no reliable decisions, therefore no results. The test is easy to run at your own company: count the tools rolled out in your stores over the past five years, then count how many are still opened every morning. The gap between those two numbers is your true team adoption rate.
In other words, the weak link was never the plan. It's the sustained, day-after-day execution, carried out by teams that keep turning over.
What 13 Years in Fresh Have Taught Us
We've been working in fresh for 13 years, across 12 countries. Long enough to have watched several generations of tools come and go, several technology trends, and several reorganizations.
That time has taught us three things.
First: in a store, what upholds the standard is almost never the system. It's someone. An assistant manager who knows their shelf, knows what's about to move, and decides fast.
Second: fresh teams are short on neither motivation nor common sense. They're short on time. And the time they spend chasing down information is time they're not spending on their shelf.
Third: a project that doesn't save time for the person on the floor doesn't survive, no matter its theoretical merits. The real judge isn't at headquarters. It's in the aisle, at 6 a.m., with two hours ahead of them.
That conviction has guided our work over the past several months. In September, we'll share what it led us to build.
One Last Question
If your systems are good and the results on the shelf still aren't showing up, the most likely explanation isn't that your systems are bad. It's that the problem isn't where you're looking for it.
The question to ask your teams this week is simple: between the moment you spot a problem on the shelf and the moment someone fixes it, how much time goes by?
The answer says more about your fresh performance than any dashboard ever could.
After 13 years of listening to fresh teams, we've built something. We're showing it for the first time at NRF Europe, September 15–17 in Paris.
Sources: Rekik Y., Oliva R., Glock C., Syntetos A., "Inventory Record Inaccuracy in Grocery Retailing: Impact of Promotions and Product Perishability, and Targeted Effect of Audits," Journal of Business Logistics, 2026 · ECR Retail Loss, "Measuring the Sales Impact of Improving Inventory Records" and "Inventory Inaccuracy in Retailing: Does it Matter?," Rekik, Syntetos, Glock · Gartner, "Gartner Predicts Over 40% of Agentic AI Projects Will Be Canceled by End of 2027," June 2025.
