ID-POS Data Analysis Using TV Commercial Viewership Data

To demonstrate the short-term advertising effects of TV commercials on shoppers, we examined the relationship between TV commercial airing and purchase timing using three types of data: TV metadata, TV viewership data, and ID-POS data. Specifically, we selected some product brands and analyzed the relationship between the gross rating point (GRP) time series and purchase timing for those brands. We selected 54 product brands in eight categories that frequently run TV commercials, including beer, carbonated beverages, instant noodles, and laundry detergents. The ID-POS data contain purchase data (over 20 million lines) for approximately 500,000 IDs shopping at approximately 70 supermarkets in the Kanto area, Japan. A no-correlation test for GRP and purchase timing using random sampling revealed a significant correlation between GRP and purchase timing for many selected brands. We further examined advertising effects from various angles by aggregating data by store, product category, product brand, and customer attributes (sex and age). The results showed some intriguing characteristics of TV commercials’ effects specific to each product brand or each customer attribute, such as the fact that older shoppers are more likely to be influenced by TV commercials, whereas shoppers at low-end stores are more likely to be influenced. In addition, by applying customer IDs to clustering, it was found that approximately 4.8% of all shoppers responded evenly to TV commercials for all brands.