Evaluation of Technical Requirements for eCommerce Methods in Traditional Retail Scenarios

Diploma Thesis by Daniel Spanagel; Carried out at Computing Department, Lancaster University and TecO, Karlsruhe University in cooperation with SAP Corporate Research (Karslruhe), supervised by Christian Decker, Albrecht Schmidt and Michael Stein.

Today a vast variety of e-commerce methods are available, which support and help the customers during their online purchase. These methods, such as product recommendations or extended information of the products, have contributed in part to the success of e-commerce companies. In traditional retail stores the customers normally don't have the possibility to access these e-commerce methods. An interesting question is whether successful e-commerce methods can be transferred to the traditional retail.

 The requirements to make such a transfer of e-commerce methods to the retail are:

  1. To identify the products of a retail store automatically.
  2. To recognise the status of the product (e.g. product status is "on the retail-shelf", "in the consumers hand", etc.)
  3. A short response time of the system. This response should be under one second as this is the time that a person can wait without losing attention.

The best solution that we could find to fulfil these requirements was a combination of two technologies: the Load Sensing Technology and the SmartShelf Technology. With this combination it is possible to detect the location and identification of an item of a typical retail product on a shelf and to recognise certain actions performed on the item, such as removing or placing from/on the shelf.  This event recognition takes less than one second. By being able to identify both the item and its status - especially if the status was "on the retail shelf" and changes to "not on the retail shelf" - we can conclude that the consumer probably has the item in the hand. Then we can use recommendations or other e-commerce methods to help the consumer in their purchase, by using, for example, an adjacent screen to display this recommendation. Furthermore, it is also possible to record information about the interaction of the consumer with such a system. For instance, if the consumer puts the product back on the shelf, there could be concluded that he/she disliked the product. If this happens often, maybe the reason could be an inappropriate package or an unwanted content.

Now, how do these two Technologies work?

The SmartShelf System works with Radio Frequency Identification (RFID). A product can be detected when a coil-antenna of the array (see Fig.1) can readout a RFID tag placed over it. From the tag, which is placed on the underside of the item, the SmartShelf can get an ID through inductivity.

The Load Sensing System consists of four load sensors and a hardware unit that process this sensor data, called Smart-Its (see Fig. 2). The four Load Sensors detect load changes on a surface placed over them when a item is placed or removed. The Smart-Its can calculate the position were the action took place over a linear relation of the load alteration.

 

Fig. 1 The SmartShelf System inside: The red squares are spots were a product can be detected inductively, by reading out a RFID tag.

 

Fig. 2 The Load Sensing System. Four Load Sensors detect load changes on a surface when a product is placed or removed. The Smart-It can calculate the position were the action took place over a linear relation of the load alteration.

Since the SmartShelf has to read out the coil-antennas sequentially, the process of recognising an event (like the removal of a product) takes too long. But, by the support of the Load Sensing System placed under the SmartShelf, it can be achieved to read out just the spot where the interaction took place, what then speeds up the event recognising process. The achieved reaction time to detect the removal of a product and its position is up to  800 ms (150-400 ms Load Sensing, and 400 ms SmartShelf). In this way it is possible to recognise a "product"-event in less than one second.

To integrate these two systems and use the combined synergy of this technologies, we implemented the following prototype (see Fig. 3):

  1. The Load Sensing System, which is placed under the SmartShelf, detects an event, e.g. the removal of an item from the shelf. This event and the position of where it took place are sent to the control and synchronization application that runs on a PC.
  2. This application maps the position to a SmartShelf spot and instructs the SmartShelf to read out thisspot.
  3. The SmartShelf sends the result of this reading process back to the PC application (a product ID of the detected item if a RFID tag was detected or nothing if there were no tag on the spot). The control and synch. application integrate the data from the SmartShelf and from the Load Sensing System.
After this process, there could be a retail scenario like the one depicted in the upper half of Fig 3:
  1. A back-end system receives the product ID and the event type.
  2. This back-end system can request a database for the associated product data to that ID.
  3. The database sends this data to the back-end system.
  4. This system adjusts the information to be shown on a display or touch screen.
  5. The consumer can request further information of the product, e.g. from the website of the product manufacturer.
 
    Fig. 3 The developed prototype and a possibly retail scenario

The results, which were achieved with this prototype, showed that it’s possible to transfer successful eCommerce methods to the traditional retail. However, there is still a high effort that must be invested in research to obtain serviceable systems for commercial applications.

Further References:

The Thesis is available in German language under "Evaluierung von Load Sensing und RFID für den Einsatz im Einzelhandel".

Contact Details:

Daniel Spanagel
Phone: +49 (0) 721 9091311
E-Mail: daniel.spanagel@gmx.de
 
Christian Decker
TecO, University of Karlsruhe
Vincenz-Prießnitz-Str.1
76131 Karlsruhe Germany
Phone:+49 (0)721 / 6902-72
E-Mail: cdecker@teco.edu