Over the past few years, Discrete Choice research techniques have gained considerable popularity in supporting the creation of optimal product and service bundles in addition to providing guidance for pricing issues.

While very effective, an oft-lamented issue is the gap between the Share of Preference provided by the analysis and the actual market shares the brand team knows to be true. And even though we researchers say over and over again that the two measures are not meant to be identical, most on the business side can’t (or won’t) separate the two.

We hypothesized that a large contributor to the difference between Share of Preference and Market Share is the respondent’s lack of real commitment to purchasing the bundle they selected in the research exercise, if it were available. So, we investigated how the “None” option is used in Discrete Choice exercises.

As part of a typical discrete choice exercise, respondents are asked to select one of several product or service offerings, where each offering contains different features. For each set of offerings, the respondent has the option to select “none” of the alternatives offered. The “None” option is asked as either:

  • “Traditional None” where it is treated as another option. In other words, if each screen includes four alternative offerings, the Traditional None option would be a fifth choice for the respondent.
  • “Dual Response None” which has become the preferred approach by most researchers is asked as two questions:
    • Which of the alternatives would they prefer?
    • If the preferred option were available, would they purchase the item?

While the “Dual Response None” tends to elicit a higher “None” than the “Traditional None”, both approaches still understate “None”.

The understatement of “None” stems from two separate issues:

  • Respondent overstatement – Consumers overstate their likelihood to purchase (a well-known challenge with survey research)
  • Lack of a valid reference point for comparison

RTi has developed a solution that addresses both of these issues and has shown an improvement in Share of Preference approximating Market Share.

  • To overcome the problem of respondent overstatement, RTi developed a questioning sequence whereby a more stringent 5-point purchase intent scale is used to better discriminate between, and thus predict, actual levels of purchase commitment.
  • To offset the issue of the lack of a valid reference point, RTi’s questioning battery refers to the product or service the respondent uses most often as a base point for helping the consumer decide whether they would, in fact, purchase the bundle they’ve been shown.

While these questioning additions might seem simple and perhaps even apparent, they have helped us to provide what we believe to be uniquely powerful Share of Preference simulators – more closely approximate Market Share than models using traditional methods of handling the “None” response.

– Howard Firestone, VP Marketing Science