To Webroom or Not: That’s the Question

The ongoing evolution towards omnichannel retail introduces novel forms of customer behaviour that are characterized by switching across channels during the purchase process such as webrooming behaviour, a phenonmenon whereby a customer searches online and purchases offline. This research examines whether online search data can predict webrooming behaviour, and aims to derive its statistical precursors by defining five different types of variables namely, (1) general clickstream data, (2) purchase funnel proxies, (3) channel interest measures, (4) historical customer behaviour, and (5) contextual factors. Using a data set that captures customer’s search behaviour online and purchase history across channels, we classify jointly the customer’s future purchase decision and channel choice. Using a combination of a state-of-the-art machine learning algorithm and SHAP values for model interpretation, we find that the historical customer behaviour, contextual factors, and purchase funnel proxies have the strongest contribution to the classifier’s predictive power. We highlight our contribution of explaining webrooming empirically which has not been done before in omnichannel retail research. The predictors from the clickstream data are the most important ones in classifying channel-wise purchase behaviour. Although our data is limited to loyalty card customers only, we are able to show the predictive value of online search data for offline purchase behaviour.