A Model for Inferring Market Preferences from Online Retail Product Information Matrices

- Shoppers navigated an online IDB clicking on cells to display attribute information.
- We model which cells are opened, how many cells, and which product is selected.
- Attribute importance weights and expected level and range of attributes are inferred.
- Data is straightforward for an e-tailer to collect.
- Results used for product, pricing, promotion, and targeting abandoned shopping carts.

This research extends information display board methods, currently employed to study information processing patterns in laboratory settings, to a field based setting that also yields managerially useful estimates of market preferences. A new model is proposed based on statistical, behavioral, and economic theories, which integrates three decisions consumers must make in this context: which product-attribute to inspect next, when to stop processing, and which, if any, product to purchase. Several theoretical options are considered on how to model product attribute selection and how to treat uninspected attributes. The modeling options are empirically tested employing datasets collected at a popular e-tailer's website, while customers were making product evaluation and purchase decisions. Subsequent to identifying the best model, we show how the resulting attribute preference estimates can be managerially employed to improve customer targeting of abandoned shopping carts for follow up communications aimed at improving sales conversions.