Your ear to the market…

Welcome to the TAILORIT blog – the platform for information and inspiration. We will continuously keep you informed about fascinating industry-specific topics, evaluate current fashion retail trends and visualize innovative store concepts. Get inspired – Change works.

Test run: “Robot vs. Human” RFID process quality

Posted on 05/19/16 by Paul Mickiewicz - Consultant

Printer Friendly and PDF

Posted on 19-05-2016 by Paul Mickiewicz - Consultant

Store personnel in fashion retail have one primary objective: to advise customers and convince them of the products. Yet often the instore logistics processes prevent them from doing just this. Thus, in the fiercely competitive apparel industry, the operational excellence of retailers is becoming increasingly important. In light of this, many retailers have decided to implement RFID. However, in order to benefit from the advantages of this technology, retailers have to regularly take stock of all RFID-articles in store using a handheld device. This presents an additional process for store personnel. When there’s a lot of activity in store, it’s possible that the cycle count process gets left out or is not completed with the necessary level of quality. This can potentially lead to gaps in inventory not being identified and the expected advantages of RFID technology not being realized. So how can one deal with this conflict of objectives?

The aim of the TAILORIT blog post “Artificial Intelligence on the Sales Floor” (Dr. Michael Baurschmid, 26.02.2016) was to provide an overview of the automated solutions that will be relevant for retail in the future. Many retailers are already looking into the use of autonomous and mobile robots. The robotics specialist Metralabs has developed a robot that is able to complete fully-automated RFID-based inventory and localization processes. For apparel retailers that already use RFID technology, the advantage of such a robot is that employees would be relieved from the additional RFID processes.

TAILORIT wanted to find out how the robot performs in comparison to store personnel in terms of detection accuracy in a real store environment. For this purpose, in 2015 practice tests with the robot "Tory" were conducted together with Metralabs in a Marc O'Polo Factory Outlet.

In the test “Robot vs. Human” the basis for comparison was the RFID items read by store personnel shortly before store closing. The robot was programmed to leave its docking station after store closing hours and to autonomously cruise through the entire store, reading the RFID-tags attached to the articles of clothing. After reading all the tags, the robot autonomously returned to its docking station. The items detected by the robot were then compared to those that had been detected by store personnel.

The first test run was conducted in May 2015. The robot exhibited a detection accuracy of 95%. The test also specifically examined who is better at detecting RFID-tagged articles located in difficult-to-access places. This included, for instance, articles that were located very close to one another in a shelf close to the floor (approx. 500 pieces) as well as articles located on metal shelves at a height of 2-3m (approx. 100 pieces). In both cases the robot was able to detect 100% of the items. In addition, localization accuracy was tested. For this purpose, RFID tags were attached to different clothing racks throughout the store. In order to test whether the robot is capable of allocating products to the correct clothing rack, samples were taken from two separate clothing racks for a total of 30 items. The test found that for 90% of these products the clothing rack assignment was correct.

The second test run took place in September 2015. This time data was collected and analyzed over the period of one week. Every night after closing time the robot drove through the store and read the articles. In the test “Robot vs. Human” the robot achieved a detection rate of almost 99%. Day to day more and more items were detected. By the end of the week the detection accuracy had risen to 99.7%. A further finding of the test run was that the robot had detected articles, which according to the system, were actually assigned to the store’s stockroom. Meanwhile, articles were also detected that were not allocated to any inventory.

What can we conclude?

  • The detection rate of the robot is comparable to that of a store employee using a handheld device.
  • A robot that regularly takes inventory will produce a more accurate recording of real-time inventory levels than store personnel irregularly taking inventory via a handheld device.  
  • Products located in difficult-to-access places can be detected without a problem.
  • The accuracy of the clothing rack localization creates new possibilities for retailers.

What do we have to watch out for?

The unique setup of each store has to be taken into account. Stores with multiple floors, steps, narrow aisles or racks with many metal parts create challenges for the robot. An additional problem is the assignment of read articles to front store and back store.

When it comes to accurately taking inventory, the robot Tory is a practical solution. Stores with a lot of retail space and lots of items are especially well-suited for the robot. The challenge now for developers is to prepare robots for particular store set-ups without a lot of effort and to find a solution as to how to separate read articles into front store and back store.

 

Photo: The robot "Tory" on the sales floor