Estimating Customer Choices – Best-Worst Scaling Vs Discrete Choice Model

The theory of economy allows us to distinguish revealed preferences (real market choices) and stated preferences. In general, we can distinguish compositional, decompositional approaches, and non-classical methods. When dealing with decompositional approaches the discrete choice method is the well-known approach in preference modeling. It allows us to model real-life behaviors. There many papers discussing best-worst scaling and discrete choice, however there is a gap in when considering delivery options. This study compares multi-profile best-worst scaling (case 3) with discrete choice methods while evaluating consumer delivery preferences in e-commerce. This paper presents an evaluation for customers in Poland. In general best-worst data can be used for discrete choice methods, if the number of profiles is sufficient. Nevertheless, by doing so we omit the information about worst profiles. Experiments indicate that delivery location has the strongest influence on consumer choice, surpassing payment method, delivery time, and shopping channel.

JEL Classification: C01, C51, C81, D12