For optimal user experience, please upgrade your browser.
Hub Bub
Loss Prevention

Can’t Wait Any Longer

Floating Widget

Floating Item Container

Floating Rate Widget




Please Select
Your Rating

This article originally appeared in STORES Magazine.

The ability to embrace different non-traditional technologies to analyze very traditional businesses like retail has real implications for high-level operational decisions.
Drew Crockett, HubBub Coffee

Surveillance video data helps retailers minimize customer abandonment

Waiting in line might be the worst part of shopping at traditional bricks-and-mortar retailers. A new system from Prayas Analytics aims to fix that, using the same video data that loss prevention departments have been gathering for decades. Those in-store surveillance videos provide valuable data about how much time customers are spending in line — and how long they’ll wait before giving up and leaving the store, taking their dollars with them.

The program is also helping to enhance store operations, particularly when it comes to queues, labor and customer abandonment around the cash register. Loss prevention, which plays a key role in helping facilitate the program, retains control over the video data but gives the program’s developers access to it.

HubBub Coffee, a coffee chain based in Philadelphia, was one of the first retailers to work with Prayas. Management wanted to see how they could optimize lines to minimize wait times.

Using data collected around queues and waiting patterns that included a complete summary of customer wait times throughout the day, HubBub learned that waits during high-traffic hours were, on average, 54 percent longer than the rest of the day. These long lines coincided with increased abandonment rates, which reached as high as 34 percent during the busiest hour of the day.

Prayas conducted a predictive analysis to determine what the expected wait time would be for various queue lengths. The recurring pattern that emerged was a 64 percent jump in wait times once a queue had more than three customers.

HubBub was able to use these numbers to quantify revenue lost in abandonments, says Prayas co-founder Yash Kothari. They were also equipped to make more informed staffing decisions based on daily queue data and projected wait times.

Drew Crockett, HubBub’s CEO, describes what he calls “the ability to embrace different non-traditional technologies to analyze very traditional businesses like retail” as having “real implications for high-level operational decisions.”

Findings like these, Kothari emphasizes, help retailers “improve customer service by refining the staffing of checkouts throughout the day.” An analysis with a large retailer piloting the product also determined that associate resources could be allocated more efficiently. The analysis discovered that 57 percent of an average associate’s time was spent with no customer in the area.

Computer vision
The technology that makes the program possible was introduced a little over a year ago by Kothari and Pranshu Maheshwari, classmates at the University of Pennsylvania’s Wharton School of Business. The co-founders are being supported by mentors on The Wharton School’s faculty, and are part of Wharton’s entrepreneur development start-up incubator program, which provides resources and mentors to a select number of entrepreneurial students.

Prayas has also received backing from First Round Capital’s student-run Dorm Room Fund, a venture firm that invests in startup companies run by students, offering young entrepreneurs mentorship and up to $20,000 in funding.

The company has received assistance from executives at retail chains including Macy’s, J.C. Penney and YUM! Brands as the team works to develop software to meet the needs of multi-store retailers.

“Bricks-and-mortar retailers are having trouble competing with e-commerce’s ability to continuously run tests,” Kothari says, “and they told us they wanted more access to data to be able to maximize their store offerings.”

Prayas’s analytical software is built on top of a computer vision platform, making it possible for a computer to “understand” the components in a video. It can tell when an object in the video is a person and can calculate when a person is moving and for how long they are moving.

The analysis “tells us what is happening in that video in terms of when a customer entered a line, how long they stayed on line, how long or when did they began speaking with an associate, how long did they talk to that associate, did they abandon a line and if so, how long was the wait before the abandonment,” Kothari says. “We do all that exclusively using the videos.”

The Prayas process is currently about 60 percent automated and 40 percent manual.  The manual component involves a team of analysts reviewing the data and the analysis to ensure accuracy, and the goal is to lower the manual rate over time.

The program can also build statistical models for enhancing labor management. “We can test each model,” he says, “to measure customer engagement and to see which model works best for any given time or day.”

Kothari points out that large online retailers like Amazon are able to do “thousands of tests on their customers and from them, figure out what an optimized online offering is.

“That’s been tremendously successful in the online world because retailers can keep running these tests, streamlining and optimizing their online offerings,” he says. “We want to bring this same quantifiable technology to bricks-and-mortar retailers.”

length of order time graph
Store Associate Engagement Graphic

Smarter staffing
As of November, a retailer piloting the program discovered that 4 percent of in-store customers abandon their purchase. The retailer, which has more than 1,500 stores in the United States and over 700 internationally, also found that customers who abandoned their transactions waited an average of two minutes and 47 seconds — nearly 90 seconds longer than the average customer.

Though it is difficult to put an exact number on revenue losses, Kothari says that initial estimates demonstrated $22 million worth of lost sales due to abandonment, as well as $5 million in losses due to labor inefficiencies.

Kothari says that the Prayas software can turn around the analysis of collected data from many stores “within a week.” The data and resulting analysis gives retailers “the ability to do smarter labor modeling, better queue [management], better employee training and better analysis and evaluation of self service checkouts.”

Down the line the data may have additional application for retailers’ loss prevention efforts — perhaps, he says, “using it in conjunction with their existing analytics to see if there are any patterns with what is happening from a customer service perspective when a theft is taking place.”

At this point, “we are working to give retailers the ability to understand things that they couldn’t see or understand just by looking at the raw data,” Kothari says. “It’s taking the data to a new and higher level, building a very strong analytical set of tools and models where we can figure out relationships between variables that wouldn’t normally hit the eye at first glance.

The Prayas Analytics system is currently looking at such metrics as:

  • How long customers wait in line
  • How quickly customers move through the line
  • How customer abandonment changes as the line gets longer
  • How engaged associates are with customers
  • What labor model works best for maximizing customer convenience
  • How to avoid both over- and under-staffing
  • How putting digital signage over or near a cash register affects customer interaction with associates
  • If digital signs affect the likelihood of line abandonment



Featured News

Loss Prevention
Episode 64 of Retail Gets Real: Theresa Payton on how to protect information online.