In

by

Content Marketing Specialist

Randa Bakken's profile picture

XeThru sensors deliver human presence tracking at store experiment

Back in September, we published the story of a client who used our XeThru occupancy sensors as part of a store experiment in the Lotte World Mall, the biggest department store in South Korea. The project, initially requested by Samsung Electronics and headed by the esteemed Hanyang University in South Korea, has now revealed some interesting and valuable measurement data that we are excited to share here.

The experiment took place over a period of five weeks, in one of Korea’s top fashion brand stores, the Beanpole store. Two kinds of measurements were set up to provide overall human presence tracking. The first was to measure how many shoppers entered the store and the other was to measure how many customers stayed in the segmented areas around the store.

The challenge for this experiment was really for the client to decide on the technology that would best fit this scenario, the one that would deliver the most precise data, without giving any false positives, nor being too intrusive. While some technologies could have been potential candidates, they all fell short when it came to convenience and performance factors. Cameras could have been an obvious choice, but the technology didn’t fit well due to customer privacy, the issue of feeling watched, and the performance level not being up to par. For a camera-based approach one would also need advanced image recognition software that comes at an additional cost and complexity. IR technology could be used for people counting, but it could not handle large amounts of people at the same time, as well as, being prone to high levels of false positives. It was concluded that radar technology is able to surpass all these challenges and deliver the necessary data needed, hence chosen over the others. The store was then segmented into nine areas and an evaluation radar module (with an average coverage area of 5mx5m) was installed per area to measure the number of browsing customers in that specific area.

Raspberry Pi's were used for data acquisition and signal processing. Each sensor node continuously sent detection data to the base station where it was processed, combined and made available. An Android application was also developed to present the data.


Raspberry Pi android app
Figure 1:Screenshot of Android application showing the number of customers throughout the week (top) and number of customers at a certain area in the store (bottom)

Main findings

Measured Entrance Data

In this first graph, we see the number of customers entering the store at every day of the week. The number clearly increases as we approach the weekend. In the below graph, we can also see there is an interesting difference in measurement data during the 2nd and 3rd week. This difference is due to the fact that the entire Lotte World Mall was on sale from the middle of the 2nd week till the Monday of the 4th week.


Entrance data
Figure 2: Graph showing the number of customers entering the store at every day of the week

Measured Sweet Spot Data

This consisted of measuring the number of people who stayed in a specific area of the store. Throughout the day, an area might get more crowded than others, hence called a sweet spot. Here we also see that the trend is increasing as the weekend approaches. Sweet spot data could indicate which collections are popular.


Sweet spot data
Figure 3: Figure showing the sweet spots around the store – most crowded areas


Sweet spot data
Figure 4: Graph showing the number of customers who stayed in a specific area of the store

It is always exciting for Novelda to see its occupancy sensors and UWB radar technology in such customer experiments and we are more than pleased when the results are successful. Stores can highly benefit from such data as they can use it to efficiently schedule staffing, review their interior layout strategy, and optimize product placement throughout the store based on customer browsing habits. This experiment proved that the XeThru technology detects human presence and movement precisely and can indeed be used in a multitude of commercial scenarios such as this one.

Acknowledgements from Novelda

Many thanks to Hanyang University and the Embedded Wireless Communications Lab led by Professor Sung Ho Cho, dragon@hanyang.ac.kr. The measurements were conducted at the Bean Pole Store in the LOTTE WORLD TOWER & LOTTE WORLD MALL. The project was requested by Samsung Electronics.

Acknowledgment from Professor Sung Ho Cho

“I thank the people in Samsung Electronics for providing me with this wonderful opportunity to challenge a pioneering research work like this. It is truly meaningful to us since nobody has done it before. I wish our work can provide additional insight and motivation in various fields of customer relationship management (CRM) applications, such as counting peoples’ movements in pathways and measuring the crowdedness in public places. We are currently extending our work to measure the massive movements of people in several subway stations in Seoul, Korea.”