Celebrities and big spenders entering Prada’s flagship store in Manhattan’s SoHo are greeted by name. For the extremely select few whose faces are well known to the public — or whose purchases are frequent enough to become store legend — consumption habits are a well-measured, real-time (private) data set. “Hello, Mr. Pitt,” whispers the black-suited salesman who appears at his side. “Always nice to see you. How are we enjoying the black cashmere scarf you purchased Wednesday?”
In less nobby retail environments — Nordstrom, say — there may or may not be someone there to greet you. Nevertheless, you can bet that the store knows you’re there. What’s more, if you’re carrying a smartphone, it might also know exactly who you are.
Online behavior is startlingly easy to track. Total traffic, IP addresses, cookies, clicks, conversions: An e-commerce site knows how long you think about buying that pair of shorts before you decide that cargos will never be in again and close the page. But the same cannot be said of brick-and-mortar stores. Retailers have very few ways of understanding customer behavior or of using it to optimize sales. Beam-break systems can count shoppers in a relatively low-tech way. Credit cards and reward card systems are often linked to a shopper’s profile. But that’s a no-go when the shopper is using cash. There is — or there was — simply no accurate means of collecting the large data sets so eagerly desired by retailers, especially in comparison with online shopping.
That’s changing. Last July Nordstrom attracted attention when the New York Times ran a story about how the upscale retailer was testing new technology to monitor the behavior of shoppers. Using technology created by Palo Alto, California–based Euclid Analytics, Nordstrom installed detection devices throughout certain stores; when customers entered, those nodes read the signals produced by the Wi-Fi cards in their smartphones. This set-up provided the company with unparalleled data: It could learn not only how many customers entered a store but also which aisles they went to and in which order, how long they considered different products — all the while knowing who they were and cross-referencing their behavior and identity with previous shopping patterns and comparison groups of shoppers.
Whereas Prada knows through old-fashioned analog data (a salesman’s attention and memory) that Brad Pitt bought a cashmere scarf on Wednesday, Nordstrom might know — through much more infallible digital data — that you stood for several seconds in front of a particular brand of scarf the last four times you were in the winter aisle. And if enough other people like you do the same, the store will likely change the placement of the product to calibrate to your behavior, such that the next time you return to Nordstrom for a pair of socks, you might inexplicably find yourself looking at that same scarf, repositioned and placed next to them.
Big-data-driven technologies by companies like Euclid and RetailNext — a San Jose, California–based provider of in-store analytics — hold forth the promise of smarter store design, more targeted selling and increased capture and margins. The benefits are many, and giddying for retailers. Imagine being able to treat a brick-and-mortar store as if it were the perfect data collection space, with its traffic as measurable as traffic online. Wireless device detection systems record actual customer behavior, whereas conventional shopper surveys rely on customer intention, which is almost always misleading.
But there are larger implications of these technologies than just more rationally optimized store layouts. Instead of deriving important economic indicators such as retail sales and consumer spending through monthly sampling based on human reporting by customers and store operators, data furnished by mobile device detection arrays could provide a real-time census of our most important economic data points. In the short term retailers with access to samples of these data could gain lead time on market-moving indicators. Over the long run, such data could be passed on or sold — directly to traders — as they are collected in real time, subverting the existing system of periodic economic indicators, incrementally collected. One can imagine a future economy that has entirely done away with lagged, sampled reporting as a means of economic data collection, and which instead has derived market indicators via streams of real-time data.
Companies such as Longmont, Colorado–based DigitalGlobe are already building and launching satellites for private sector analysis firms like RS Metrics in Chicago that use the imagery — mixed with sophisticated statistical and sampling techniques — to count cars in the parking lots of Walmart and Target stores, as proxies for their customer traffic. Eventually such efforts will be linked together, and cold war–style satellite surveillance will be integrated with in-store customer counting and measurement technologies to form an aggregated, real-time, continuous signal of local and national consumption patterns that can be streamed to Wall Street analysts 24 hours a day.
As market-moving information arrives faster and more continuously from real-time sources, markets will actually move more gradually and gently. The practice of eagerly lingering around a screen at precise times (like 8:30 a.m.) of precise days, jittering about for a drip feed of economic data — dispensed by the federal government — seems almost hypnotic when one takes a step back. Yet it is nevertheless the way capital markets have learned about major adjustments in macro assumptions about growth and spending since the beginning of electronic communications. All that will soon change.
The experience of receiving key economic data once a month or even once a week is jarringly disconnected from the reality of a continuous, 24-hour-a-day stream of actions and interactions that make up the real-time economy (almost all of which can be measured, sampled or approximated in some way, on the fly). Our present reliance on such intermittent signals of economic health results in discontinuous jumps in information (such as major downside or upside surprises, which were in fact gradually, continuously — and measurably — forming by the minute and day). These needless surprises then have to be abruptly discounted by markets when they become known, often resulting in brief but intense periods of quaking volatility in the minutes and hours after new economic indicator data are released.
Reading about this system of economic data measurement will one day seem as quaint to future traders as is our impression of physical sheets of listed equities doled out by exchanges to traders lining up in the morning outside physical buildings.
Read more about financial technology.
