Choice Modeling to Screen New Product Concepts

Creating viable new product concepts is a slow, expensive, and failure-prone process. Often the process begins with focus groups and/or depth interviews, sometimes followed by ideation and brainstorming, to develop or stimulate the development of new product concepts. Then, the nascent product concepts go through managerial screening, and lastly, on to traditional concept screening and testing among consumers. Is this really the best and most cost-efficient way to develop and evaluate new product concepts?

The Starting Point

Choice modeling is often a better and more cost-efficient way to generate and screen hundreds (or even thousands) of unique new product concepts. Choice modeling refers to a family of tradeoff and conjoint techniques. These techniques are primarily used for optimization and forecasting. To illustrate how choice modeling might be used, let us assume that the goal is to develop new barbecue sauces. The goal, new barbecue sauces, is the starting point, and the starting point is entirely a human decision, a judgement call. Now we can go to work.

Foundational Research

The first step would be a major secondary research investigation of failed and successful barbecue sauces, including reviews of barbecue sauce recipes published in books and magazines in the past. This secondary research would also include reviews of historical and current data on the barbecue sauce category from all sources. The next step would be some type of qualitative research (focus groups or depth interviews) to help identify potential barbecue sauce variables and the ranges of those variables. Let’s suppose that our research (secondary plus qualitative) identifies the following variables:

Type of smoke flavors (4): hickory, mesquite, white oak, red maple
Sweetness level (4): very sweet, sweet, slightly sweet, no added sweetener
Sweeteners (4): honey, brown sugar, maple syrup, corn syrup
Sets of spices/flavors (4): allspice-cinnamon-cloves, coriander-cumin-paprika, chili powder-garlic powder, onion powder-jalapeno
Bits (3): onion bits, bacon bits, celery bits
Price levels (3): low, medium, high

If we pick one element from each of the above variables (underlined), the total number of unique combinations is 2,304. So, with one choice modeling project, 2,304 new product concepts could be screened and the relative market potential predicted for each concept. The cost per concept would be a small fraction of what concept screening typically costs.

Choice Modeling

Here’s how choice modeling works. A mathematically derived experimental design is created, a subset of the 2,304 possibilities. Then, each target-market consumer would evaluate six to 10 scenarios. Think of a scenario as a shopping trip. The survey participants would see one new product concept and be asked to indicate how many jars of that new barbecue sauce, if any, would be purchased in the next 30 days, if the new product were available where the consumer normally shops. That completes scenario one. Then, all of the variables change following the experimental design, and the participant is asked to indicate how many jars of the new barbecue sauce they would buy. This completes scenario two. Then, all the variables change again, and the respondent repeats the exercise.

Typically, each study participant completes 6 to 10 scenarios or “shopping trips.” If the total sample size were 1,000, then the final analysis would be based on 6,000 to 10,000 scenarios. The experimental design allows analysts to derive equations to predict the overall appeal of each of the 2,304 new product concepts, even though only a subset of the concepts is evaluated. This is the magic and the cost-efficiency of choice modeling.

This very simple example of choice modeling to screen new product concepts is merely to illustrate the basic idea. Many different types of choice models can be used, and other variables can be incorporated into the research design. For example, competitive brands could be included in the design so that market shares could be predicted for each new product concept. Also, variables and/or values of variables can be omitted or hidden (if a particular combination makes no sense).

Consumable Goods and Durable Goods

While this example is focused on barbecue sauces, choice modeling can be used for a wide range of consumable products (i.e., consumer packaged goods) and durable goods (trucks, appliances, chainsaws, power tools, etc.). Sometimes, surveys about durable goods are more challenging because consumers might not be familiar with the specific terminology or be aware of the latest technical developments. In these cases, some type of video demonstration might be needed to make sure participants understand the features and the descriptive language that explains those features. If a durable good is complex (e.g., an automotive vehicle), it might be necessary to have a car clinic up front so that study participants can learn all of the variables and terminology, and then the choice modeling experiment would follow.

Decision Simulator

The ultimate deliverable from choice modeling is a Decision Simulator. The equations derived from the choice modeling experiment are used to build an interactive simulator, so that “what if” possibilities could be explored. What if the price of a concept could be lowered? How would that affect its likelihood of success? What if the spice level was dialed up? What if sweetness levels were varied? What ingredients could be substituted if an ingredient is in short supply? Hundreds of possibilities could be explored to find optimal solutions. The Decision Simulator is reliable for a year or two after its creation, so it can help guide new product decisions for a significant period of time.

A Powerful Tool

Regardless of the product category, choice modeling offers a powerful set of tools to systematically evaluate a much wider range of new product concepts (compared to traditional methods) at a much lower cost per concept. Choice modeling also reveals the relative importance of each variable in explaining preference for a given new product concept, an understanding that can lead to more and better new product concepts.

Author

Jerry W. Thomas

Chief Executive Officer

Jerry founded Decision Analyst in September 1978. The firm has grown over the years and is now one of the largest privately held, employee-owned research agencies in North America. The firm prides itself on mastery of advanced analytics, predictive modeling, and choice modeling to optimize marketing decisions, and is deeply involved in the development of leading-edge analytic software. Jerry plays a key role in the development of Decision Analyst’s proprietary research services and related mathematical models.

Jerry graduated from the University of Texas at Arlington, earned his MBA at the University of Texas at Austin, and studied graduate economics at SMU.

Advanced Analytics Blog