STUDY ON SPECIFYING EFFECTIVE KEYWORDS FOR GOOGLE SHOPPING ADS -USING AD IMPRESSION DATA AND SOCIAL MEDIA INFORMATION-

Abstract

ABSTRACT The penetration rate of Internet advertising in Japan has risen to over one trillion yen by 2014. There are various kinds of Internet advertisements, and Google Shopping Ads are one of the representative ones, that recommend appropriate products for words (words in a search query) searched by consumers. To provide appropriate products and information for a search phrase, appropriate keywords should be included in product descriptions and titles. However, it is not easy to specify appropriate keywords. In this study, we aim to specify appropriate keywords for search terms in Google Shopping Ads by using social media information. Specifically, we first identify topics related to the fashion domain using data collected from Twitter. Based on the topics identified, we create a discriminant model to identify the search terms that contribute to conversions for fashion in general and for each product category. As a result of these analyses, we have identified noteworthy topics that include trendiness for each product category. Keywords Social Networking Services, Natural Language Processing, Topic Model, Random Forest, Google Shopping Ads, Ad Impression