Choice-based Conjoint and Discrete Choice Models: Version 4

Room 2 | 10:20 am - 10:50 am | Tuesday, October 11, 2022

Discrete Choice Models combined with Conjoint Analysis – Choice-Based Conjoint Analysis (CBCA) – are a mainstay of quantitative marketing research. Three eras of innovation each improved upon prior tools to yield better explanations of behavior: aggregate choice models (V1), Latent Class Choice (V2) and Bayesian Individual-Level Models (V3). This presentation will introduce the latest, fourth generation of CBCA covering the estimation of the buyer’s budget in addition to price sensitivity, uncovering which attributes people pay attention to when choosing and statistical alternatives to the standard estimation tools that are ubiquitous in CBCA.

In the 1970s, Daniel McFadden developed Discrete Choice Models to understand the drivers of transportation choices, earning him the 2000 Nobel Prize in Economics. Paul Green introduced Conjoint Analysis in the 1970s to study the trade-offs between brand, price and features when rating products and services.

In the early 1980s, the brilliance of Jordan Louviere combined these into CBCA, realizing that trade-offs are made when choosing product and services. All these analyzed aggregates of people rather than the choices of groups or individuals. These define CBCA V1. The mid-1980s saw Latent Class (LC) CBCA uncovering segments of people who made similar trade-offs. I co-authored the first academic paper introducing LC CBCA. Moving from aggregates to segments define V2 of CBCA. In the late 1990s, Bayesian models quantified the trade-offs of individuals. CBCA V3 has been dominant for the past twenty years. No significant advances have been developed since then.

We introduce three innovations that are the harbingers of CBCA V4:

1. While CBCA estimates price sensitivity, it does not tell us anything about the buyer’s budget, which constrains how much they can spend. Budget Constrained CBCA can estimate the budget constraint for the choice of a single product or the choices from a menu of features or products.
2. CBCA assumes that everyone considers all brands, features and prices when making a choice. Yet we know that many people will simplify and pay attention to just a subset. Attribute Attendance CBCA uncovers who pays attention to which attributes under study.
3. Finally, that standard statistical model has certain built-in assumptions about how people make choices. We describe alternate models that make other assumptions about choice behavior.

The dominant feature of these V4 innovations is simple: better explanation of choice behavior than standard V3 CBCA.

Key takeaways:

1. Understand how estimation of a budget constraint changes estimated price sensitivity and attribute importances.
2. See how different people pay attention to different choice drivers, rather than attending to all drivers and how that result changes predictions.
3. Realize that the standard choice model may not produce the best fitting results in every situation and therefore other statistical approaches should be considered.

Topics covered:

• New techniques – qualitative and quantitative.
• Predictive analytics.

Methodologies Presentation by Supplier