Subscriber Discussion

Thermal Cameras In Kroger Supermarket?

I'm assuming by "high-tech infrared cameras" they're talking about thermal, in this Consumerist article (links to a WSJ article that's behind their paywall).

Funny enough, the writer's "mom and pop grocery store" anecdote works equally well in large stores... I know at most of our local supermarkets, if things start to back up, a cashier will hop on the phone and put out a public announcement for reinforcements. Seems to work well enough, as another register usually opens within a minute or so.

Maybe today's crop of Kroger cashiers aren't as quick-thinking, though...

Here's some details from the WSJ article:

"Kroger's system, dubbed QueVision, is now in about 95% of its stores, which operate under the Kroger, King Soopers and Dillons names, among others.

The system includes software developed by Kroger's IT department that predicts for each store how long those customers spend shopping based on the day and time. The system determines the number of lanes that need to be open in 30-minute increments, and displays the information on monitors above the lanes so supervisors can deploy cashiers accordingly.

Kroger declined to specify the cost for installing QueVision, but says the cost of running it is minimal. While it is hard to break out the exact impact faster lines have had on sales, the company says surveys show customer perception of its checkout speed has improved markedly since 2010."

I watched QueVision's video and I am more confused now than before. Infrared detectors not cameras are used. Like REX detectors on an access control system.

Doing some digging I found this:

Irisys’s queue management system includes a “people counter” above each entrance/exit that tabulates the number of people entering and leaving the store by using heat-sensitive infrared sensors. The system also features infrared sensors above the checkout lanes that can detect the number of “units” of shoppers — a unit could include a mom and two kids — in line, and their average wait time.

The system then crunches the data gleaned from the sensors to predict the number of lanes that should be open in 15 minutes, and in 30 minutes, to maintain the retailer’s customer service standard at the checkout. (This could mean opening or closing lanes.)

Read More: Supermarket News profile on Kroger

Good find, thanks for sharing that article, very information. The company they cite (Irisys) and the setup they describe actually sounds very expensive.

Matt hits on something that I've always questioned about retail analytics... what is their real value??

I do not dispute that retail analytics can technically accomplish exactly what they say they can (which is a very rare utterance regarding analytics in general). However, so what?

Each and every one of the things that retail analytics say they can do, can be accomplished (and has been for decades) by 1 or 2 semi-alert manager/employees.

Can the 'machines' do it better than humans? Without question. Unless there is a supermarket Rainman somewhere...

And what about the equations they use to justify costs? Nobody has an issue with the math?

The 'projection numbers' the analytic companies use (i.e. how much stores that are smart enough to see the value of analytics can 'save') are only valid if the employees ignore noticing these same traffic-flow data! In order to calculate any increase, you have to establish a constant reference base (how much the store is losing before analytics were rolled-out). This has to remain constant for any future projections to be valid.

Roll out some kind of reward program for your employees that gives them an incentive to notice traffic flow patterns - and save your chain of stores tons of loot. Loot you were previously losing (however nebulous you might find that calculation to be), plus all the loot you save not buying stuff you don't need to solve a relatively simple problem.

Someone PLEASE clue me in to what I am missing?

Well, you wanted to know who was using them. Now you know!

I shop at Kroger all the time. Next time I'm going to bring a ventriloquist dummy equipped w/a heater/blower attached with me to mess with their analytics.

I sure hope Kroger wasn't sold on this technology by watching this.

^^^ That gets a 9.5/10

They seem to relish cheeseball marketing?

Here's a more mainstream Kroger commercial. Clearly, some suit at Kroger HQ drank the quevision koolaid and rolled it into an entire 'customer satisfaction' campaign based on lowering the customer's 'average speed to checkout' :)

I am with Marty on this one... a cannon when a fly swatter is is all that is needed.

I worked at a Wegman's grocery store as a young man for 4 years and one of the primary duties of the 'front end supervisor' was to make sure there were not any long lines for the customers.

If there were... folks would get called up front to man the registers. They trained anyone they thought could do the register job just so they could pull this off.

This worked great and only needed a pair of eyes an someone who payed attention.

The Kroger system deploys Irisys thermal counters, which connects in real-time to the POS system in order to achieve reasonable predictive scheduling (how many cashiers to open in the next 10, 15, 20 minutes) and is considered a big marketing success for promoting customer service. Currently, there are only a few frontline management roll-outs, including ASDA (Wal-Mart UK) by Brickstream with video counters, Morrison's and Kroger by Irisys, and partial roll-out in Tesco (Irisys). Due to a marketing push by Irisys and Brickstream, there is a lot of interest by American supermarkets and in 2013 will probably see a pile of pilots.

Frontline management (or queue management for main banks) is not simple and requires expertise both in technology and deployment, but there are many advantages. Think in terms of having a security camera instead of a security guard; a 24/7 empiric system instead of objective observation by individuals. The more obvious benefit is customer perception, but there is also the optimization of active cashiers, real-time deployment, and real-time monitoring (instead of mystery shoppers), which generate money saving opportunities. Moreover, there is the 'soft' business benefit of using hard core data as basis of communication between corporate and the local store.

Long story short, video analytics has many benefits, from monitoring sales conversion, building schedules based on service intensity and service productivity metrics, in-store staying and browsing behaviors for marketing, queue management and predictive scheduling for the checkout process. Each market - retail, big box, supermarkets, airports etc. - has different requirements but the technology has evolved enough that we can start doing really interesting behavior analytics.

Ronny, thanks for the detailed feedback. Any idea of the price of doing this per store or line? Also, any idea of why thermal cameras where chosen over conventional surveillance cameras? I presume superior performance / accuracy of the thermal ones?

Hi John, there is a story behind why Kroger choose Irisys, suffice to say that today there are two market leaders in frontline management - Irisys with thermal and Brickstream with video. In addition to the eternal debate of thermal vs. video, there are two factors to consider in the solution.

First there is the queue sensor. While traffic counter mounted at the entrance to the store measures 'motion', the queue sensor must be able to deal with 'standing' behaviors without releasing the object and therefore losing the count. This is has profound implications to correctly measuring waiting time, which, to me, is the key to an optimized solution. The installed and calibrated sensors range around $1500 to $2500, depending on the number of tills they cover and their accuracy, including the ability to deal with 'abnormal behaviors'.

Second, the frontline solution requires expertise in both queue technology and predictive scheduling. The challenge in costing the solution is not only in deciding how many sensors are required and where to position the cameras, but also in the analytical, real-time application that can determine how many cashiers should be active in the next 10, 20, 30 minutes. The ball park cost for 8 tills store is between $10K to $25K depending on the components of the solution, the requirements for real-time connections with POS and WFM systems, and the number of stores.

Rony, thanks, very insightful!