What is Choice Modeling?

Consumers can tell us what they like and do not like. They can tell us what they will buy and what they will not buy. But rarely can they tell us why they buy one brand over another. They do not know the roles that price, brand image, package color, brand name, promotional offers, and media advertising play in their purchase decisions. But with controlled scientific experiments and advanced multivariate analyses, we can implicitly measure the marketing variables that underlie the consumers’ purchase decisions.

Discrete choice modeling, volumetric choice modeling, and conjoint modeling are analytical methods used to simulate real-world consumer purchasing behavior. Our Advanced Analytics consultants set up carefully controlled experiments in which consumers are simply asked to choose how many of each product they would buy, given predetermined sets of realistic conditions. Consumers are not consciously aware of what is being measured. The brands are presented visually (including 3D simulation), if possible, in the context of advertising, pricing, packaging, features, promotion, and other variables. The importance of each marketing variable is then derived mathematically.

The “choice” experimental design is tailored to the specific objectives, constraints, and variables of the project. Customization is the key to success because every category/brand has critical idiosyncrasies.

Choice modeling techniques can help marketing researchers:

• Analyze price sensitivity
• Bundle product and service features
• Optimize brand strategy
• Improve product-line planning
• Maximize media advertising effectiveness
• Improve promotional offers
• Optimize advertising messages
• Improve package designs

Conjoint Analysis

Conjoint analysis is ideal for optimizing new product designs by identifying the most appealing sets of features. Conjoint analysis, sometimes referred to as trade-off analysis, is a multivariate technique that quantitatively measures the relative importance of different marketing variables, attributes, or product features related to a brand, product, or service. The distinguishing feature of this technique is that each variable’s importance is determined implicitly or indirectly. That is, the respondent is not fully aware of what is being measured.

Discrete Choice Modeling

Discrete choice modeling is ideal for (a) product categories where only one purchase is made over a longer period of time (for example, durable goods, credit cards, cellular phones, etc.) and (b) complex products (i.e., products with many different possible features). In these carefully controlled experiments, current and potential customers are asked which one product they would buy, given a realistic scenario including all of the products or services that compete with one another in the marketplace. In each scenario, the respondent is presented with a different set of marketing stimuli and asked which brand or product would be purchased. The type of decision that the respondents make in each scenario is designed to mimic the real market, and again each variables’ importance is being determined implicitly.

Volumetric Choice Modeling

Volumetric choice modeling is ideal for product categories where (a) multiple products are purchased over relatively short periods of time and (b) repeat purchase volume is an important consideration. In these experiments, current and potential customers are asked how many of each product they would buy, given a realistic scenario including all of the products or services that compete with one another in the marketplace. The type of decision that the respondents make in each scenario is designed to mimic the real-world marketplace, where varying quantities of multiple brands might be purchased (including volumetric measures such as units purchased, dollars spent, etc.). The overriding goal is to create realistic scenarios that properly represent the buying behavior that we are striving to model. And the modeling is designed to implicitly measure the role and importance of each marketing variable.