Where Choice Modeling Is Applied: Four Relationships That Shape Decision-Grade Evidence
Choice modeling is applied across consumer packaged goods, technology, software, financial services, insurance, healthcare, transportation, and many other industries. But simply listing industries does little to explain why choice modeling works differently across applications.
In choice modeling consulting, the challenges differ across industries because the relationships underlying purchasing and adoption decisions are not always the same. A consumer choosing between competing brands, a business renewing a software subscription, and a hospital evaluating medical equipment face fundamentally different decision problems. Decision-grade evidence requires understanding the relationships among alternatives, over time, with context, and among decision-makers.
These four relationships provide a useful framework for understanding where choice modeling is applied and what makes its application challenging.
Four Relationships in Choice Modeling
- Relationships Between Alternatives: How products, services, and configurations compete with one another, including substitution patterns and shared product characteristics.
- Relationships Over Time: How current choices interact with subscriptions, renewals, future benefits, and ongoing customer relationships.
- Relationships With Context: How individual circumstances, prices, locations, availability, and other constraints shape the alternatives people can realistically choose.
- Relationships Among Decision-Makers: How individual preferences contribute to decisions made by organizations, households, committees, or other groups.
These are not four separate choice modeling methodologies. They are four types of relationships that may need to be represented when applying choice modeling to real decisions.
1. Relationships Between Alternatives
Choice modeling is frequently used to understand competition among products and services. Applications include product design, pricing, portfolio optimization, and understanding substitution among alternatives.
Consumer Packaged Goods (CPG)
CPG is among the most established applications of choice modeling. Researchers examine how consumers trade off brands, product characteristics, and prices when selecting among competing products. Many CPG applications are relatively straightforward. However, certain problems require more careful treatment.
Price-Pack Architecture (PPA)
Price-pack architecture introduces a distinctive modeling challenge because different pack sizes may contain exactly the same underlying product.
Consider a two-pack and a four-pack. The four-pack contains twice the quantity, and the alternatives are closely related. If the four-pack is discounted, substantial substitution from the two-pack should be expected. A standard multinomial logit (MNL) model that treats each pack as an unrelated alternative does not automatically recognize this relationship. Its substitution assumptions may therefore produce unrealistic predictions. Price and pack size can also become correlated within the experimental design, complicating estimation and interpretation.
The challenge is not simply estimating which pack consumers prefer. It is representing the relationships between packs in a way that produces coherent substitution patterns.
Technology, Durable Goods, and Automotive
Technology products and durable goods frequently involve large numbers of relevant attributes. Computers, appliances, and other sophisticated products may require consumers to evaluate many characteristics simultaneously.
The challenge is representing those characteristics without assuming that a large number of attributes is inherently a problem.
Automotive research adds another complication: the enormous universe of available products and configurations. Even restricting an exercise to a particular vehicle category can leave a difficult number of alternatives to represent.
2. Relationships Over Time
Some choices are relatively discrete transactions. Others initiate ongoing customer relationships. For these applications, understanding the initial choice may be only one part of understanding its economic consequences.
Software and Software as a Service (SaaS)
Software companies frequently offer subscription plans with multiple pricing tiers, feature bundles, and different levels of functionality. Customers may also evaluate subscriptions based partly on anticipated future benefits and functionality. The resulting choice problem combines immediate feature and price trade-offs with expectations about an ongoing service relationship.
Telecommunications
Telecommunications shares many characteristics with software subscriptions. Consumers select service plans, bundles, and contract arrangements that may continue or change over time. The initial plan selection and the subsequent customer relationship are related but distinct economic considerations.
Financial Services
Financial services can have particularly complex revenue relationships. A product or service may be free or inexpensive to the customer while creating value through transactions, complementary products, network effects, or a deeper financial relationship. Consequently, the economic value of acquiring a customer may not be captured by the price or revenue associated with the product initially selected.
Insurance
Insurance introduces two additional considerations. First, premiums are frequently individualized according to risk. Two customers may face different prices for essentially the same coverage. Second, insurance contracts often renew automatically. Actual customers may continue their existing coverage without actively comparing alternatives, while choice experiments invite respondents to make repeated selections. These differences matter when interpreting predicted switching and the consequences of pricing decisions.
3. Relationships With Context
In some applications, the alternatives available to a person depend heavily on individual circumstances. The relevant choice set, prices, and perceived convenience may differ from one respondent to another.
Airlines
Airline choices depend on origin, destination, travel dates, schedules, and route-specific prices. A common experimental scenario may not closely resemble the flights a particular respondent would actually consider. Yet tailoring scenarios to each respondent’s travel requirements introduces considerable complexity. The difficulty is representing realistic choices when the available alternatives and their prices depend on the travel context.
Transportation and Mobility
Transportation choices are strongly influenced by individual circumstances. Convenience, in particular, can mean very different things to different people. The same transportation option may be practical for one respondent and unsuitable for another.
Healthcare and Public Policy
Healthcare decisions may involve treatment attributes, patient circumstances, access, and competing outcomes. Public policy applications may involve choices about transportation, energy adoption, environmental programs, or public services. In these settings, individual constraints and the consequences of choices can be especially important to interpreting the evidence.
4. Relationships Among Decision-Makers
Many choice experiments measure the preferences of an individual respondent. But in some markets, the individual being interviewed is not the person who independently determines what gets purchased.
Business-to-Business (B2B) Products and Services
Organizational purchasing frequently involves multiple stakeholders with different responsibilities and priorities. A technical specialist may value product performance. Procurement may emphasize price and contractual terms. Management may be concerned with organizational or financial consequences. Measuring the preferences of any one participant does not necessarily explain the organization’s eventual decision.
Medical Technology (Medtech)
Medtech provides a particularly clear example. A physician may have strong preferences for a particular device, but a hospital’s purchasing decision may also involve procurement, administrators, and other stakeholders. A choice model based on physicians’ responses may accurately represent physician preferences without fully representing the institutional purchasing decision.
The underlying issue is the relationship between whose preferences are measured and who actually determines the choice.
Choice Modeling Applications by Industry
The four relationships provide a common framework, but individual industries may involve several relationships simultaneously.
| Industry or application | Common choice modeling applications | Particularly relevant relationships |
|---|---|---|
| Consumer Packaged Goods | Brand, product, and pricing decisions | Between alternatives |
| Price-Pack Architecture | Pack sizes, price structures, substitution | Between alternatives |
| Technology & Durable Goods | Features, configurations, product design | Between alternatives |
| Automotive | Vehicle selection, features, configurations | Between alternatives, with context |
| Software & SaaS | Subscription tiers, functionality, pricing | Over time, between alternatives |
| Telecommunications | Service plans, bundles, contracts | Over time |
| Financial Services | Products, fees, customer propositions | Over time |
| Insurance | Coverage, premiums, policy selection | Over time, with context |
| Airlines | Fares, routes, schedules, service options | With context |
| Hospitality | Accommodation, amenities, pricing | Between alternatives, with context |
| Transportation & Mobility | Travel modes, convenience, services | With context |
| Healthcare | Treatment and patient trade-offs | With context |
| Energy & Public Policy | Adoption, incentives, public services | With context |
| B2B & Medtech | Equipment, services, institutional purchasing | Among decision-makers |
| Higher Education | College and MBA program selection | Between alternatives |
This is not an exhaustive classification. Many industries share choice structures, while individual applications within the same industry may require different considerations.
From Industry Applications to Decision-Grade Evidence
Choice modeling methodologies can be applied across many industries, but the existence of a statistical model does not guarantee that its results adequately support the intended decision.
A price-pack model may estimate preferences while misrepresenting substitution between closely related packs. A subscription model may explain initial plan selection without establishing the economics of the ongoing relationship. An airline choice experiment may estimate preferences for alternatives that do not realistically represent a respondent’s available choices. A B2B study may measure individual preferences without adequately representing organizational purchasing. These are different problems, but they illustrate the same principle.
Decision-grade choice modeling requires the evidence to reflect the relationships that matter for the decision being made.
The appropriate methodology, experimental design, model specification, and interpretation depend on the application. Not every industry requires a different model, but different applications may require attention to different assumptions. Understanding where choice modeling is applied is therefore only the beginning. The more important question is whether the analysis represents the economic and behavioral relationships necessary to support the decision.
For related discussions, see Common Methodologies in Discrete Choice Modeling and Choice Modeling Consulting.
