Assessment of Long - Short Term Relation Between Category Sales and IoT Sensors: A FMCG Retailer Application


Eskiocak D. I., Oral F., Mert B., Zeybek O., Kilercik B.

4th International Conference on Intelligent and Fuzzy Systems (INFUS), Bornova, Turkey, 19 - 21 July 2022, vol.505, pp.480-487 identifier identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 505
  • Doi Number: 10.1007/978-3-031-09176-6_55
  • City: Bornova
  • Country: Turkey
  • Page Numbers: pp.480-487
  • Keywords: Internet of Things, Business analytics, Retail analytics
  • Istanbul Technical University Affiliated: No

Abstract

Developing communication technologies have led to a significant paradigm shift in the retail sector. Many processes that used to be supervised by a human controller have been automated and become manageable via sensors. Today, many operational tasks in stores are regularly monitored by IoT devices, and necessary arrangements can be made instantly. This study examined two stores of one of Turkey's leading FMCG retailers. Time-series data is aggregated into 30-min intervals of temperature, humidity, air quality and light sensor readings from devices placed in the aisles. These measurements are matched with sales data corresponding to the category sales of the shelves in which IoT sensors are installed. The Autoregressive Distributed Lag (ARDL) method from the Econometric Time Series (ETS) family of models analyses the relationship between the ambient conditions in the aisles and the sales. These models act on the assumption of long-term co-integration between sales and sensor data. Results point out that the sensor readings have a significant explanatory effect on sales. The application of the co-integration test revealed that there is a long-term relationship between sensor data and sales. The long- and short-term effect models estimated in this direction present that the ambient values have a statistically significant effect on sales.