At the Quirks Event in London today, Jon Puleston of Kantar discussed the challenges with using traditional personality tests in commercial market research. Ogilvy now emphasizes cognitive segmentation rather than demographic segmentation, as advertising has shifted from demographic targeting to microtargeting. At the behest of Ogilvy, Kantar profiled 30,000 panelists around the world using the OCEAN Big 5 Personality Test. This led to the following realizations:
- Classical tests fail to differentiate consumers. Crosstabulating the results to surveys by personality revealed few differences.
- The tests varied dramatically by country. Looking at the tests across countries produced dramatically different assessments by dimension; for instance, Australians reported as having much lower openness than the French.
- The Big 5 measure is incomplete. Knowing the common five facets of personality didn’t provide other important measures about outlook and decision-making. Ogilvy had looked at the personality of people wanting to lose weight and found the attitudes didn’t match the hypothesis and the personality measures were too generic for the context of people who wanted to lose weight.
The Big 5 measures are “useful in theory, but useless in practice,” Jon said.
Of course, these personality tests were not designed specifically for international commercial market research and as a result had shortcomings for such work:
- Limits of self-assessment. The heart of the issue is that classical personality tests are very reliant on self-assessment measures with high levels of confirmation bias. While respondents had no issue reporting neutral personality aspects (such as introversion vs. extroversion), few would admit that they are “unconscientious, closed-minded, or disagreeable.”
- Overreliance on repetitive Likert scales. Use of Likert scales was subject to acquiescence response style, with a favoring of the top two items for scales; for instance, 35% overlap in people who selected “Moderately Accurate” or “Very Accurate” to both “I am chatty/talkative” and “I am quiet and reserved.” This affect was made even worse for respondents using smartphones instead of PCs or tablets. Worse, attention spans started to fade going through the Likert scales.
- Lack of commercial usefulness. Questions about emotional stability and attitude to risk, among others, aren’t grounded in consumer decision making. An emotional range in personality didn’t differentiate between anxiety (who might research best brand) and emotional (who might rely on feelings to choose a brand).
- Uncontextualized generic questions. Being asked if you are “conscientious” in general provides little predictive power for consumer purchases, which might range from impulsive (a candy bar at checkout) to considered (a new home).
These shortcomings led to iterative development of new surveys and eventually resulted in a proven, 17-minute questionnaire suitable for profiling panelists. While this instrument is proprietary, Jon shared some lessons:
- Replace Likert scales with select-all-that-apply personality labels. A list of competing choices reduced overlap between mutually exclusive choices (e.g., from 35% for chatty/reserved using the Likert scale to 8% using this approach).
- Use the Silent Dog method to reverse items like “I am untidy” (12%) to “I like tidying up” (30%, implying 70% are untidy).
- Broaden self-assessment to include others. While few admit to being quarrelsome, more will admit to being more quarrelsome than their brother.
- Use a retrospective perspective. Reframe self-assessment to be of the past self. “I’m not the type of person who is late to meetings, but I was often late to school.”
- Use memes. Just by showing the image of the Grumpy Cat to respondents, twice as many people admitted to being grumpy.
- Provide behavioral context. Only 12% would admit to being disorganized, but 50% admitted that the clothes in their bedroom drawers are disorganized. Multiple similar questions can provide a better assessment of whether someone is disorganized.
- Reframe the context to that of consumer decision-making. Instead of asking if people agree “I am organized” ask “I make a list of things I need before going shopping”, instead of “I am not a risk taker” ask “I always check the labels when I go shopping.”
Based on these findings, Kantar wanted to go beyond the traditional Big Five personality measures. Kantar fielded other tests:
- Boyd Past, Present, and Future
- Hedonic vs. Utilitarian
- Locus of Control
- Need for Affect
- Need for Cognition
- Regulatory Focus
- Zimbardo Time Perspective
For social issues, Kantar used a Yale measure of a worldview of hierarchical vs. egalitarian, individualistic vs. communitarian.
Through pilot experiments, iterative testing, principle component analysis, and international fielding, Kantar and Ogilvy developed their personality profiling tool. Due to the overlap of measures, by refactoring the tests they were able to streamline a 30-minute questionnaire to 17.5-minute questionnaire. To encourage concentration and completion, the survey provides feedback throughout the survey after about every 3 minutes of answers.
Kantar then tested the tool across a range of studies:
- Measuring the appeal of advertising by personality
- Devising better anti-smoking messages
- Understanding the barriers to getting mothers to vaccinate their children
- Understanding the barriers to losing weight, using a cognitive segmentation
- Understanding the personality of a company’s sales organization and its customers.
Personality profiling was very predictive of whether someone smoked or not, and smokers typically avoided negative emotions. As a result, smokers skipped past negative ads, clicking past these ads 30% faster than non-smokers. Taking thinking style into effect, new ads provided positive emotional messages focused on the external reasons for giving up smoking (for their family). Smokers spent 50% longer viewing these ads than nonsmokers.
Personality profiling can produce business impact in studies, provided the researcher reaches beyond academic personality tests.
ESOMAR members will want to download the full paper for this presentation; this presentation won the Best Paper award at ESOMAR Congress 2019.