We are happy to announce, that we will have 3 researchers from all 3 chairs presenting during the next PhD seminar, Tuesday 26th October.

Giulia Crestini and Radu Tanase will present together and inform us about their project:

Title: 
Does Pricing Transparency Benefit or Harm the Customer-Retailer Relationship? – A retailer and consumer perspective 
Abstract:
Over the last years and also due to the COVID-19 pandemic, consumers increasingly shop online mainly due to lower prices, convenience, and access to customer reviews (GlobalWebIndex 2020). To identify lower prices, consumers use price charts displaying a product’s price history, typically provided by price comparison sites but recently also by online retailers. While prior research found that price charts influence consumers’ purchase timing, evidence from the field and of their impact on consumers’ perceptions of the retail brand is yet missing from the literature. This research examines whether introducing price charts displaying a product’s price history on online shops can be cause of harm for the online retailer (e.g. through increasing consumers price sensitivity) or instead could positively affect consumers’ search and purchase behaviors (e.g. reducing information asymmetry).
–     –     –
In addition, Joel Persson, who started recently at Florian von Wangenheim’s Chair at the ETH will also present:
Title:
Off-Policy Learning of Dynamic Content Promotions
Abstract:
We present an off-policy learning framework for optimal dynamic content promotions on webpages. Our method is motivated by a marketer (or online editor, content creator) that sequentially in time must determine which of the currently relevant content (e.g., “editors picks”) shall be promoted on each of several owned and curated distribution channels (e.g., website homepage, email newsletter, social media platforms) such that viewership, readership and engagement is maximized. The policy is constructed by, for each channel, first learning the causal uplift of promotion, then learning which covariates are prescriptive of uplift heterogeneity, and finally, at each point in time ranking content with respect to the predicted uplift given the prescriptive covariates. We partner with a leading national newspaper to test the performance of our method in practice. Counterfactual evaluations on historical data shows that our method outperforms current editorial practice and baseline methods in terms of policy value. Our research contributes to marketing in two ways: First, by developing a framework for off-policy learning of optimal dynamic content promotions on the web across multiple distribution channels. Secondly, by empirically demonstrating the effectiveness of such policies compared to alternative methods and manual practice.