Content uploaded by Matthias Hauser
Author content
All content in this area was uploaded by Matthias Hauser on Feb 17, 2017
Content may be subject to copyright.
13th International Conference on Wirtschaftsinformatik,
February 12-15, 2017, St. Gallen, Switzerland
Leveraging RFID Data Analytics for the Design of an
Automated Checkout System
Matthias Hauser1, Sebastian Günther1, Christoph M. Flath1, and Frédéric Thiesse1
1 University of Würzburg, Department of Business Management, Würzburg, Germany
matthias.hauser@uni-wuerzburg.de, sebastian.guenther@aol.de
christoph.flath@uni-wuerzburg.de, frederic.thiesse@uni-wuerzburg.de
Abstract. Traditional checkout systems are labor-intensive and can be a source
of great frustration for customers having to wait in line. In contrast, automated
checkout systems seamlessly scan, total and charge a customer’s purchase to a
registered payment account while they are simply leaving the store. We focus on
the main challenge of automatically detecting customer purchases. To this end,
we develop a checkout system that leverages data mining techniques to
(i) reliably and timely detect items leaving the shopping floor area and (ii) assign
them to individual customers. We demonstrate the system’s feasibility using a
large data set collected in the laboratory under real-world conditions.
Keywords: Data Analytics, RFID, Internet of Things, Automated Checkout
1 Introduction
While RFID solutions are commonly used in retail supply chains for automatic
detection of logistical units in upstream and backroom processes, front store customer-
facing applications have not progressed beyond pilot trials. RFID solutions for this “last
mile” of the supply chain are especially challenging because of the sheer number and
variety of simultaneously moving objects. Complexity is further increased by the
manner in which objects are transported (stacked, in bags, etc.), unpredictable customer
behavior, and suboptimal store layouts. Against this backdrop, we consider RFID-based
automated checkout systems, which promise great value potential for the retail industry
[1]. Our research objective is to decompose the challenge of such systems into tractable
sub problems and demonstrate the feasibility of a pilot implementation.
2 Automated Checkout System Requirements
Traditional checkout systems are labor-intensive and can be a great source of frustration
for customers when having to wait in line. To reduce costs retailers have started
adopting self-service systems which enable shoppers to scan, bag and pay for their
purchases with little or no help from store personnel [2, 3]. These systems, however,
offer hardly any improvements over traditional checkout with respect to the customer
1201
Hauser, M.; Günther, S.; Flath, C.; Thiesse, F. (2017): Leveraging RFID Data Analytics for the Design of
an Automated Checkout System, in Leimeister, J.M.; Brenner, W. (Hrsg.): Proceedings der
13. Internationalen Tagung Wirtschaftsinformatik (WI 2017), St. Gallen, S. 1201-1204
experience, potentially offering an even worse experience. In contrast, automated
checkout systems scan, total and charge a customer’s purchases to a registered
payments account while they are simply leaving the store.
In our research, we focus on the main challenge of automatically detecting customer
purchases. Here, we must cope with two problems: The system (i) has to reliably detect
all purchased products and (ii) assign these to individual customers. Undetected
products cause direct losses to the retailer and inventory inaccuracy resulting in out-of-
stock situations [4]. Incorrectly assigning items to a customer, on the other hand, may
lead to customer dissatisfaction and interruptions of in-store operations [5].
Academic literature on automated checkout systems is sparse. To the best of our
knowledge, only one group of researchers have developed a system (“MyGrocer”) that
addresses the previously mentioned challenges [6, 7]. This solution relies on shopping
carts equipped with RFID readers that detect objects placed in the carts. Similar to this
approach, we also rely on RFID for the detection of products. In contrast to optical
barcodes, RFID allows identification of products at the item level. In addition,
identification does not require a direct line of sight between the tag and the reader
device, which allows for simultaneous bulk detection of multiple objects. In contrast to
the previous studies, we decided against RFID-equipped carts for several reasons: First,
shopping carts are not used in all types of retail stores (e.g., fashion retail stores) which
limits the generalizability. Second, the costs for equipping shopping cart with RFID
readers are very high [7]. In addition, there are high operating costs for regular service
of the carts. Instead, we propose an automated checkout system that is implemented at
the exits of retail stores and does not require customers to use specific devices. This
system facilitates the detection of items in carts and items carried by customers.
The MyGrocer carts only need to detect items within them. In this case, sufficient
data quality can be achieved by selecting RFID antennas with read ranges of only 20cm.
In our case, the antennas need to detect items that leave a store through an exit gate,
which requires antennas with large read range and high power. Unfortunately, this in
turn leads to the detection of RFID tags carried nearby by the gate instead of through
the gate. Secondly, assigning items to customers can hardly be realized with hardware-
based approaches. Customers would have to wait in line and pass the gate one after the
other to avoid purchases of different customers being read at the same time.
3 Description of the Automated Checkout Artifact
We pursue a design-oriented research approach to create an automated checkout artifact
that (i) reliably and timely detects items leaving a store and (ii) correctly assigns them
to individual customer. The architecture combines hardware components and software
components. The hardware component consists of two RFID reader installations, a
ceiling-mounted system which tracks items in the store and a gate-mounted system that
helps to detect items that are leaving the store (see Figure 1). This infrastructure collects
low-level RFID data that is then processed by the software components.
1202
Figure 1. Hardware infrastructure with two distinct RFID reader installations
There are two distinct software functionalities. The first software component uses data
mining techniques to reliably and timely detect items that are leaving the store. To this
end, we extract features from the low-level RFID data, which contain information
regarding observed real world events. One simple example is the maximum received
signal strength value measured in a series of detections of a particular tag within a
certain time interval. With these features we train classification models that are able to
distinguish between items that leave the store through the exit gate and others. We
follow the conceptual approach described in Hauser et al. [8]. However, we go beyond
the state-of-the-art by considering run fragments in real-time instead of evaluating
completed paths ex-post. The second software component assigns items leaving the
store (identified by the first component) to individual customers. To this end, we infer
item paths in the shopping area and then apply cluster analysis to group them. The
procedure rests on the assumption that the paths of items purchased by one customer
are more similar to each other than to paths of other items.
4 Expected Contribution and Future Work
We use a large data set collected in a laboratory environment for instantiation and
evaluation of the artifact. Our experimental setup takes into account the limited process
control in retail stores by considering multiple walking paths, varying numbers of
persons and items as well as different movement speeds. Our experimental design
includes 18 different tests. We repeated each of the tests 50 times. We then perform 5-
fold validation to ensure representative results. We found that the artifact reliably and
timely detects all shopping baskets carried out of the shopping floor area. On average,
the baskets were detected 0.41 seconds before customers stepped out of the sales floor
area. In addition, 94% of the shopping baskets assigned by the artifact contained the
correct items.
These misclassifications arise in two particularly challenging test scenarios where
multiple customers approach the exit gate simultaneously on very similar (almost
RFID gate with reader and gate-
mounted antennas
Shopping area
Ceiling-mounted RFID reader with 52
far-field antenna beams
Store exit
1203
identical) movement paths. In practice such a situation could easily arise when friends
are shopping together which highlights the limitations of the pilot implementation.
Nonetheless, our initial study demonstrates the fundamental feasibility of RFID-based
automated checkout. Going forward, we want to evaluate our artifact in different test
environments to evaluate its generalizability. In addition, we want to include additional,
richer test settings (e.g., scenarios in which customers take objects from shelves that
are placed near the exits). Our ultimate objective is to ensure feasibility under real-
world conditions.
5 Acknowledgement
This research was carried out under the Research Project SERAMIS (Sensor-Enabled
Real-World Awareness for Management Information Systems), funded by the EU
under the VII Framework Research Program.
References
1. Manyika, J., Chui, M., Bisson, P., Woetzel, J., Dobbs, R., Bughin, J., Aharon, D.:
The Internet of Things: Mapping the Value Beyond the Hype. McKinsey Global
Institute (2015).
2. Orel, F.D., Kara, A.: Supermarket self-checkout service quality, customer
satisfaction, and loyalty: Empirical evidence from an emerging market. J. Retail.
Consum. Serv. 21, 118–129 (2014).
3. Litfin, T., Wolfram, G.: New Automated Checkout Systems. In: Krafft, M. and
Mantrala, M.K. (eds.) Retailing in the 21st Century: Current and Future Trends. pp.
143–157. Springer Berlin Heidelberg, Berlin, Heidelberg (2006).
4. Kang, Y., Gershwin, S.B.: Information inaccuracy in inventory systems: stock loss
and stockout. IIE Trans. 37, 843–859 (2005).
5. Hayes, R., Blackwood, R.: Evaluating the effects of EAS on product sales and loss:
Results of a large-scale field experiment. Secur. J. 19, 262–276 (2006).
6. Kourouthanassis, P., Roussos, G.: Developing Consumer-Friendly Pervasive Retail
Systems. IEEE Pervasive Comput. 2, 32–39 (2003).
7. Roussos, G., Kourouthanasis, P., Spinellis, D., Gryazin, E., Pryzbliski, M.,
Kalpogiannis, G., Giaglis, G.: Systems Architecture for Pervasive Retail. In:
Proceedings of the 2003 ACM Symposium on Applied Computing. pp. 631–636.
ACM, New York, NY, USA (2003).
8. Hauser, M., Zügner, D., Flath, C.M., Thiesse, F.: Pushing the limits of RFID:
Empowering RFID-based Electronic Article Surveillance with Data Analytics
Techniques. In: Proceedings of the Thirty Sixth International Conference on
Information Systems (2015).
1204