In 1943, Abraham Wald was handed a question about bombers returning from missions over Europe. The aircraft were coming home with bullet holes scattered across their wings, fuselage, and tail, and the instinctive conclusion was straightforward: reinforce the areas taking the most damage. Wald, working with the Statistical Research Group at Columbia University, saw that the problem lay not in the analysis but in the sample. The aircraft being examined were the survivors. Any plane hit in a genuinely critical area had never returned to be measured, which meant the damage on the bombers that made it home showed only where an aircraft could be hit and still survive. The undamaged areas were not safe. They were unexamined, because the evidence about them was sitting at the bottom of the English Channel. Wald's work on aircraft survivability, later documented in the Journal of the American Statistical Association, became a classic illustration of what happens when a model is built only from the cases that survived the process being studied.
Retail has a version of the same problem, and it is hiding in plain sight inside every board report. Once a customer completes a transaction, the volume of information a retailer holds about them is extraordinary: the point of sale knows what they bought, what they paid, and when; loyalty data connects that purchase to previous ones; basket analysis reveals which products travel together; and average transaction value, units per transaction, and sales per square foot describe the commercial outcome in detail. Yet the customer who walked in, spent fifteen minutes browsing, and left with nothing disappears from that picture almost entirely, as does the person who looked through the window and kept walking, and the shopper who arrived during a rush, could not find help, and gave up.
The point of sale sees none of this. No transaction means no record. But the business still lost something: a customer who came in and left without buying. They are the planes that never came back.
Your Next Read 👉 Why Traffic Counting Alone Is No Longer Enough
Sales Data Counts Buyers. Traffic Data Counts Shoppers.
Every dataset a store generates is triggered by a purchase. Transaction records, basket composition, loyalty enrollment, and margin mix are all created at the moment someone says yes, which makes them precise descriptions of completed journeys and silent on abandoned ones. Retail has spent decades becoming exceptionally good at measuring successful customer behavior while treating unsuccessful behavior as an absence of data rather than as data in its own right.
NielsenIQ described a version of this problem in August 2026 in a report called The Invisible Shopper: What Retail Data Isn't Telling You, arguing that retail data is very good at answering what shoppers bought from us and close to silent on what shoppers did everywhere else. The in-store version of that blind spot is narrower and more immediate. It is not only that a customer bought elsewhere. It is that they stood inside your store and bought nothing, and no system recorded that they were ever there.
E-commerce never developed that blind spot, because online the denominator is impossible to miss. A digital team can see sessions, product page views, search activity, and cart abandonment, so conversion became a natural operating metric from the beginning. A physical store has always had exactly the same denominator walking through its doors. It simply has not always had a reliable way to measure it. That single difference explains far more than it appears to, because a business that can only count outcomes can only grow by producing more outcomes. It opens more doors, extends more hours, carries more inventory, and schedules more staff, because you cannot optimize a rate you have never computed.
This is precisely the gap Traffic Analytics is built to close. By measuring who passes the storefront, who enters, how many people are inside, and how those movements track against sales, it gives the transaction a denominator and turns store performance from a total into a rate that can actually be managed.
Revenue Rankings Measure Location, Not Store Performance
RetailNext's own data from Black Friday weekend 2025 shows how far apart traffic and sales can travel. Across the Friday to Monday period, store traffic fell 5.8% year over year and net sales fell 4.4%, which on its own reads as a straightforward decline. Yet average unit retail rose 4.4%, average transaction value rose 2.1%, and shopper yield, the average sale generated per person entering the store, edged up 0.3%. Fewer people came, and retailers extracted slightly more from each one who did. A revenue figure on its own cannot separate a demand problem from an execution problem, and those two call for opposite responses.
The same ambiguity sits inside every store ranking. Consider two stores: the first turns over $2.1M on 180,000 visits, the second $1.4M on 85,000. On revenue alone, the first is the flagship and the second is the problem, and the promotion and the performance plan get handed out accordingly. Divide by traffic, though, and the first extracts $11.67 from every person who enters while the second extracts $16.47, which is 41% more from each customer received. The quieter store is not the weaker operator. It is the better one, working a weaker trade area. Rank a portfolio by revenue and you learn which stores have the best locations. Rank it by how much of the available opportunity each captures, and you learn which stores are being run well.
That second ranking is only possible once traffic is measured consistently across every location, which is what the Benchmarks Hub reports from actual sensor data across thousands of stores each month rather than from surveys or modeled estimates.
Knowing that a store is converting below its potential is the beginning of the question rather than the end of it, which is where Insights takes over. Full path analytics, dwell, and heat mapping show where inside the store shoppers are engaging and where they are disengaging, so a conversion gap can be traced to the zone, display, or journey actually causing it.
Learn More 👉 How Traffic Data Accuracy Affects Labor Costs And Conversion
Only 36% Of Store Associates Say Staffing Matches Actual Traffic
In a March 2025 survey of 500 frontline retail associates across the United States, only 36% said staffing schedules consistently align with actual store traffic. More than half, 51%, reported being short-staffed during busy periods most of the time. Nearly three-quarters, 73%, had watched customers leave because nobody was available to help them, and 77% said their store regularly loses sales because of scheduling decisions. A parallel survey of 500 UK associates in July 2025 found 84% observing lost sales from poor shift scheduling, while 68% also saw overstaffing during slow periods. The problem is rarely too few people in total. It is people scheduled against the wrong hours.
That gap exists because most schedules are built from last year's sales rather than from this month's expected demand, and sales only tell you when people bought, never when they arrived. Knowing how many people visited yesterday explains yesterday. Knowing when demand will arrive changes what happens next month. RetailNext's predicted traffic forecasts demand in 15-minute intervals up to 45 days ahead using each store's own history, which puts expected traffic in front of the labor plan while the schedule is still being built rather than after it has been locked.
Forecasting only matters, of course, if the underlying count can be trusted, and traffic is rarely an endpoint. It becomes the denominator in conversion, the basis for staffing, the reference for store comparison, and an input into marketing and store planning decisions, which means a small measurement error does not stay small. It travels through the system and changes the conclusion at the other end. That is why every Aurora® sensor installation is audited against blurred video from a randomly selected date and time before its counts are used at all, reconfigured and re-audited until it clears the threshold, and re-verified annually thereafter. At the point where traffic data determines whether a store appears to have a conversion problem, or whether labor should be moved, it has stopped being a metric and become infrastructure. Infrastructure has to be trusted.
Once it is, the questions come faster than dashboards can answer them, which is what Pulse AI exists for. A regional manager who sees conversion slipping can ask which store is driving it, whether traffic changed there, and how staffing compared with last month, in plain language and in sequence, without building a report between each question.
Measuring Store Traffic Reveals The Customers You Already Have
UNTUCKit is a useful illustration of what changes, precisely because the company began online and already knew how to run a rate business. Its digital operation could see the whole journey toward a purchase. Its stores could see only the transaction at the end of one. Once store traffic was measured and connected to conversion, the company could compare traffic patterns against sales, evaluate each location against the number of shoppers it was actually receiving, and make operational decisions on something better than a final transaction count, producing a reported 6.9% increase in conversion, an 18% increase in shopper yield, and a 40% reduction in IT infrastructure costs. The UNTUCKit customer story matters because the technology did not create that traffic. The customers were already arriving. The measurement simply made their behavior visible.
This is ultimately what Wald's example has to do with retail. The problem was never that the available data was inaccurate. It was that the process generating the data determined what could be seen at all. A transaction database describes customers who bought with real precision, and explains customers who considered buying, entered and left, or approached the storefront and kept walking, not at all. Those customers have not stopped being commercially relevant. They have only become analytically invisible, and a shopper you cannot see is an opportunity you cannot assign, measure, or reward anyone for capturing.
So the most useful question is probably no longer how many people came into the store. It is what happened to everyone who came close to buying and did not. Once a retailer can answer that, traffic analytics stops being a counting exercise and becomes a way of recovering the part of the customer journey that transaction data erased.
Keep Reading 👉 Why Enterprise Retailers Choose RetailNext
