Finding Closure Around Store Closures
It’s been a rough stretch for Sears. The term ‘store closure’ has almost become synonymous with the brand. The big box retailer announced that it would close the least profitable locations on seven separate occasions last year, reducing its footprint to 400 locations.
When we began tracking the giant in 2011, the chain was composed of over 12,000 locations and our analysts manually marked each parking lot so we could begin monitoring its traffic. Read more here on how and why we use car counts as a way to analyze a retailer’s health. But what happens when retailers open and close brick and mortar locations? The short answer is that we’ve invested heavily in the best process to ensure real-world changes in daily store locations are reflected in our retail product. Now for the long answer…
A few definitions
Before we dig into the technology behind sourcing, maintaining, and continually updating a growing database of 300,000 stores, below are two key definitions you need in order to get from pixels—little dots that comprise an image—to insight, which to us, is actionable intelligence.
Geofencing: A set of longitude and latitude coordinates, drawn as a polygon, that mark an area of interest (AOI), like a Walmart parking lot.
Metadata: Location-specific information like the store’s opening date, closing date, and hours of operation. This critical dataset allows for accurate point in time data attribution.
How we do it
Welcome to the retailpocalypse. Its latest victims include Sears (and Kmart), Bon-Ton, and Toys ‘R Us. But you don’t have to file for Chapter 11 and liquidate to close a store location. Retail locations close every day due to local taste changes, site selection strategy, and corporate decisions to shed unprofitable locations. For example, Macy’s, JCPenney and Chipotle all closed locations last year but are not in danger of bankruptcy.
Unfortunately, there is no official data source for store closures, so we use multiple third-party sources to continuously comb for stores that have closed. While “web scraper” data vendors like AggData enable us to scale this research to accommodate a database of 300,000 AOIs (which translates to 300,000 individual parking lots) we still lack a way of being alerted when a store closes. We think of knowing exactly when a location opens and closes as a “time dimension” and you can think of it as a store’s lifetime.
For instance, a Starbucks customer can easily find a new location on Yelp, but that’s not enough for us. We need to know the exact day that Starbucks location opened and then continually track its lifecycle to know if the store moves or when it eventually closes. By monitoring locations at this granular of a level, we are able to create accurate “point-in-time” datasets. One way we automate this process is by connecting our AOI database to Google Maps and matching each location with a unique Google Places ID. Thanks to Google, locations are updated with the frequency needed for accurate traffic attribution.
While headlines generalize retail as “overstored” and in a “corrective phase”, retailers are still opening locations. Since we cover approximately 95% of the locations for each chain in our coverage, adding the newest Dollar Tree to our database of 20,000 DLTR locations won’t significantly alter our correlations. However, knowing when stores open improves our accuracy over time, especially in smaller chains where store growth represents meaningful change.
When we add new tickers to our universe, having reliable store opening data is crucial. We recently added The Container Store, Noodles & Co, Ollie’s Bargain Outlet, and Sprouts Farmers Market to our consumer product, so it was critical to understand all of the new store openings for each of those tickers.
By continuously sourcing, maintaining and updating our database of 300,000 retail locations, we benefit from sampling more locations, improved correlation to revenues and growing ticker coverage. Just how important is metadata? We’ve found that updating locations once a quarter results in at least +/- 1% change in over 30% of observations.
Updating store closures in our database isn’t as exciting as real life store closures. However, having accurate store data still a critical input in our analytics designed to quantify and measure social, economic, and environmental activity on Earth, and that’s what we’re here to do.
To find out more, contact email@example.com