Potential buyers of a product or service consider many factors when deciding what to buy, and potential sellers must decide on the factors to offer and at what price. Buyers typically place their own economic self-interests first: spend the least amount of money, time, and effort to get the product or service they want. Companies want to satisfy customer needs but also have a competing objective: to find the factors that maximize appeal, revenue, and profit.
The general class of research designs that deals with intersecting (conjoined) variables in a structured manner is called conjoint analysis (covered in a previous post). Conjoint analysis involves reaction to, and statistical relationships between, factors in choice decisions. A “factor” is part of the choice criteria used to make a buying decision. A factor also has characteristics, features, or benefits that vary called “levels”. Simple examples of factors are price, quantity, and size. Some real-world examples might include:
- An industrial paint sprayer company might offer a unit with multiple hose lengths, paint capacity, nozzle diameter, or spray wand length.
- A new credit card with different levels of an annual fee, number of airline points per dollar spent, levels of travel protection coverage, or access to airport lounges.
- A laptop manufacturer may offer different screen sizes, keyboard layouts, processor speed, graphics cards, memory, hard disk size, or bundled software.
In each of these examples there might be just one or two levels per factor, or perhaps many more. In its entirety, one set of combinations (factors x levels) could be rendered as a series of concept description (e.g., known as a “full profile” design) typically used in new product development and optimization. In other cases, such as competitive preference testing, options are shown as sets of choices to respondents (e.g., choice modeling or discrete choice designs). We then obtain demand estimation through intent to buy or intended volumetric consumption.
As complexity increases (more factors with more levels), the number of possible combinations of products grows exponentially. For example, four factors with three levels each produces 34 = 81 unique combinations. Theoretically, we could show all 81 combinations to a single respondent – but would we want to? Poor quality data and exhausted subjects are two good reasons not to.
Conjoint analysis solves for the problem of too many combinations by showing each subject a randomized subset of all possible combinations, also known as a partial factorial design (i.e., a subset of the complete factorial design). Features or characteristics of a product or service are sequentially exposed using concept descriptions and simple explanations. In most studies we recommend a basic explanation of the exercise to familiarize respondents with the category and tasks. Concept features and levels then vary by subject in the overall matrix of combinations. In simplistic terms, this akin to Swiss cheese: we know what the overall shape is (complete design), but there are holes inside (partial exposure).
In conjoint analysis we strive for balanced “orthogonal” designs in which an equal number of factor x level combinations are shown to a respondent, usually as a set of even numbers (i.e., 8, 12, or 16). As expected, more complex conjoint designs require much larger sample sizes for statistical reliability. From this we interpolate demand curves for each factor (e.g., price, size, quantity) and also assess overall demand for the optimal combination of factors that maximize consumer appeal. In addition to demand curves for each factor, and identification of the optimal combination of factors, additional outputs include simulation and revenue estimation for the optimal set (there may be multiple) of product configurations.
Deficiencies of Conjoint Designs
Conjoint designs assume that:
- Screening criteria accurately reflect the true target market.
- Respondents are consistent and rational in decision making.
- Alternatives are meaningfully different (subtle differences may produce no test effect).
- Realistic combinations of factors and levels are always shown (non-confounding design).
- Buyers and buying decisions seek maximum benefits at the lowest price. This may not be true depending on a category, such with overwhelming altruistic or social cause benefits.
- Price has linear and important, but not overwhelming, influence on choice. Respondents may quickly become sensitized to pricing which may exaggerate true differences.
As you might tell from the above, the design and the screening criteria, as well as the factors and levels chosen for exposure, are a huge factor in making sure that the conjoint results ultimately make sense. A simulator provides scenario testing using different combination of features and using the utility scores that were generated from the respondent sample. Conjoint utility scores are converted to a scaled value (in the above case, a 65% interest level). When applied against an interested target or segment, an estimate of penetration can be developed. Sales estimation would need to be confirmed with additional forecasting and analysis (an additional stage of work).
From the same research study, a max-diff analysis of individual features was performed (i.e., based on preference ranking). Scores are preference “votes” based on a series of choice exercises, ranked from high to low. You can see from this slide that higher paint transfer efficiency (through the nozzle), ease of operation, higher paint transfer efficiency (from container), reduced mist, and less likelihood of bursts or droplets were the top items. Mechanical function items were lower on the preference hierarchy
This snapshot attempted to show the range of conjoint designs and approaches that could be used depending on a client’s needs. Each individual design is based on the considerations of the business decision that needs to be made and the number and complexity of choices to be evaluated by respondents.
For more information, please contact:
Bob Walker, CEO
Surveys & Forecasts, LLC
https://safllc.com
+1.203 255.0505