Editor's note: Norman Frendberg is president of Consumer Insights, Rochester, New York.

Creating a research design that accurately addresses the marketing issues at hand represents an important and challenging role for the research professional. This challenge is further enhanced by the wide variety of appropriate alternative research designs available, for virtually every marketing issue. In some cases, the difficulty increases when marketing management specifically requests a particular design which may be considered less than optimal.

For instance. a marketing manager may ask the researcher for help in understanding the causes of a specific consumer behavior by conducting a survey only among a particular population. For example:

  • Interview consumers who responded to a promotion in an attempt to understand its success/failure.
  • Interview heavy users of a product to understand what drives their consumption.
  • Interview consumers after trial of a new product formulation to determine

One-shot studies

The above examples use a design which is often referred to as "one-shot studies,' involving tile collection of data from only one sample group of respondents.

This design may be appealing because of its simplicity and low cost, but it lacks a control cell for comparison of the key measures. Subsequently, such an approach may yield very little insight about the behavior under study.

For example, the maker of a major brand of ground coffee wants to determine consumer acceptance of a new formulation (i.e., the mixture of Colombian, African, and Asian coffee beans). Let's assume that marketing management directs a one-shot study by testing the product with a sample of consumers. Furthermore, the findings indicate that 60% of the respondents rated that product "excellent" overall.

The following key question still remains unanswered: are consumers positive to this new coffee?

By restricting the sample population, we have not determined if consumers are positive or negative to this new coffee mixture. In order to accomplish this, we must compare the 60% "excellent" new product rating to a relevant control cell product, which might be the current mixture.

Popular designs

A product test which easily incorporates an overall rating measure with a control product can he executed in several ways. The two most popular designs are the monadic and the protomonadic.

In the monadic design, two equivalent but different respondent samples are interviewed. Each respondent tests either the control product (i.e., current) or the new formulation test product, but not both.

The protomonadic design involves interviewing only one set of respondents, with each respondent testing both the control product and the new formulation test product, but in a sequential format. The first product is tested and evaluated, followed by the testing and evaluation of the second product. (In practice, half of the respondents test the new formulation product first with the current second, while the other half of the sample tests the products in reverse order. This rotation eliminates any order bias, i.e., preference for the first product tested, regardless of which one.)

Both research methods are very strong designs since the scores for both the test and control products are attributed to the actual product differences.

Let's refer back to our hypothetical example where 60% of the respondents rated the test product as "excellent." The examples below illustrate two different control cell results which would totally alter our conclusions.

Situation 1

Control (current)
Product measure


Test product

60% excellent

60% excellent


Conclusion: The test product is at parity with the current. If the new formulation is less expensive to produce, then it should be introduced.


Situation 2

Control (current)
Product measure


Test product

80% excellent

60% excellent


Conclusion: The test product fails to exceed the current and should, in most cases, be eliminated from consideration.

Although the situations may be extreme, they exemplify two vastly different conclusions with the same post-only results (i.e., 60% "excellent" rating).

The control cell design is regularly used for the type of product testing described above. However, some research projects without a lengthy history or an acceptable method of product testing may lack a control cell. Regardless of acceptability, a control cell is just as important when conducting a promotion test because without this control group for comparison, we fail to have sufficient data to provide accurate assessment of the promotion.

Stairstep to disaster

Selecting an appropriate control product is not always as straightforward as using the current formulation. Let's assume that marketing management continues to reduce the cost of the coffee product formulation (e.g., adding inferior coffee beans at the expense of superior ones). If we continue to use each subsequent "current" product, i.e., already cost-reduced, as the control cell, then we could experience a phenomenon known as the "stairstep to disaster."

The "stairstep to disaster" process occurs because each new product formulation is only slightly inferior, which fails to be significantly less preferred than the current product. However, if several small degradation steps are taken, the resulting product (i.e., at the bottom of the stairstep) can be quite inferior to the original formulation.

While product testing of the "current" versus the "cost reduction formulation" is appropriate to test one new formula, some control is needed to avoid this stairstep to disaster when many new cost reduction formulas are involved. Several approaches exist for creating an appropriate control product to be used to test the latest cost reduction formulation. One method entails the creation of an original formula to test against each new cost reduction formula, rather than testing the new product against its predecessor. Another approach involves testing the new cost reduction formula versus one or two competitors in the category who deliver similar overall benefits to the consumer.

Summary

In brief, a control cell product is needed in order to answer absolute questions that marketing management often requires such as "How good is this new product or promotion?"

Without a control cell, we cannot convert our survey answers (i.e., 60% "excellent") to insights about consumer acceptance of our product.

The selected study design must be able to answer the key questions asked:

  • What have I really learned?

  • How conclusive is the evidence?