I have spent 25 years digitizing the physical world, from barcodes and RFID to automation systems, helping enterprise customers turn physical operations into data they could manage. Every generation of technology got better at capturing information. But the fundamental flaw never really changed: it all depended on humans to collect it accurately, interpret it correctly and act on it in time.
The original barcode was invented because one out of every 32 keystrokes at a cash register was a statistical error. We have come a long way from that. But if you spend time with retail operators today, you hear a version of the same frustration: we have more data than ever, and we still cannot act on it fast enough to change what happens next.
The industry has a term for data that is clearly wrong or missing. That gets flagged. What is harder to deal with is data that looks right in dashboards and quarterly reports but falls apart the moment someone on the store floor needs to make a real-time decision. I have started calling this “nearly usable” data. And it is quietly one of the most expensive problems in retail.
What Nearly Usable Data Actually Costs
Nearly usable data does not announce itself as a problem. It is not a system outage or a failed integration. It is subtler than that. An alert fires at self-checkout, but the associate does not trust it, so she overrides it to keep the line moving. A compliance gap shows up in the fresh department, but by the time the markdown recommendation reaches the floor, the window to recover value has closed. A report tells the regional manager that shrink is trending up, but it arrives 48 hours after the behavior that caused it.
None of these are dramatic failures. Each one is small. But they compound across thousands of stores, hundreds of thousands of transactions, every single day. The cost does not show up as a single line item. It shows up as margin erosion that is maddeningly difficult to diagnose.
This is why shrink has moved from loss prevention meetings to the CFO’s agenda. Executives are realizing how much recoverable margin is leaking through operational gaps their data should be catching, but is not. When your CEO scrutinizes your fifth-largest IT line item every year and asks what more they are getting for it, “we generate good reports” is not an adequate answer. The question is whether the technology is actually recovering revenue.
The Data Puddles Problem
People in this industry love to talk about data lakes. In my experience, most enterprises do not have data lakes. They have data puddles: pockets of information scattered across systems that were never designed to talk to each other.
Point-of-sale data lives in one system. Inventory sits in another. Video footage is archived in a third. Workforce scheduling is somewhere else entirely. Each puddle has value on its own. But the operational decisions that actually protect margin, such as staffing the pharmacy when the line gets long, pulling product before it expires or intervening on a loss pattern before it becomes a trend, require connecting signals across puddles in real time.
Most organizations cannot do that today. And AI does not solve this on its own. I share the excitement about agentic AI and autonomous decision systems. But I have deployed AI across enough enterprise environments to know that the first hurdle is always the data puddle problem. If the signals are fragmented, delayed or missing context, you simply accelerate confusion.
From Reporting Losses to Recovering Revenue
One of the clearest signals from recent industry events is that retail leaders are done talking about loss prevention. They want to talk about revenue recovery.
Loss prevention is backward-looking. It counts what walked out the door. Revenue recovery asks: how early did we detect the problem, could we intervene before the margin was lost, and did the system make it easier or harder for the associate to do the right thing?
I have heard this directly from customers: “Do not call what you do loss prevention.” You undervalue what you do when you talk that way. They are right. When you reframe the problem as revenue recovery, the conversation moves from the asset protection office to the CFO’s office. It becomes a P&L discussion, not a shrink report.
And this reframing extends beyond traditional theft. Fresh waste is increasingly recognized as preventable shrink. When produce spoils because the markdown came too late, or a pharmacy queue builds and customers leave because staffing did not adjust, that is recoverable margin lost to the same underlying problem: data that arrived too late to change the outcome.
What Decision-Grade Data Actually Requires
Decision-grade data has to be timely enough to change an outcome, not just explain one. It has to be grounded in something observable, something a store manager can look at and say, yes, that is what actually happened. It has to reduce the cognitive load on associates, not add to it. And it has to produce a shared version of reality that the frontline and the C-suite can both trust.
That last point is where most retail data falls short. Executives see one set of numbers. Store managers see another. Associates see alerts they have learned to ignore. When there is no shared evidence anchoring the conversation, every level of the organization defaults to its own interpretation, and alignment breaks down.
The Question that Matters Now
The retailers I see pulling ahead are not the ones deploying the most tools. They are the ones asking a harder question: of all the data we collect, how much of it is actually usable when it matters? How much of it can our associates trust and act on? How much of it can our CFO tie directly to recovered margin?
If the honest answer is that most of your data is “nearly usable,” then that gap is where your margin is leaking. No amount of new AI layered on top will close it until the underlying signals are trustworthy, timely, and actionable.
Retail does not have a data shortage. It has an execution gap. The companies that close it will not be the ones with the most dashboards, but the ones whose data can still change the outcome when it matters.
Joe White is the CEO of Everseen, a leading Vision AI platform for retailers. He previously served as Chief Product and Solutions Officer at Zebra Technologies, where he oversaw strategy, investment and product development across the company’s extensive solutions portfolio. White joined Zebra in 2014 following the acquisition of Motorola Solutions’ Enterprise Business and held senior leadership roles at Motorola, Motorola Solutions, CAIS Internet and Digex. He is passionate about driving the adoption of enterprise technologies and is a strong supporter of those who serve in the U.S. Armed Forces.
