Fun Facts About Concept Testing

Purchase Intention: Predicting success in concept screening

Companies spend millions of dollars every year using purchase intention as a way of predicting success in concept screening. Given the reliance on “purchase intention” as a key predictor or success, there probably isn’t as much scrutiny of the measure as there should be.

So here it is.

On average, only 28% of people who said they would “definitely or probably try” a product, actually did. Only 40% of people who said they would “definitely try” the product, actually did so. These observations result from multiple PhD validation studies. They show that while purchase intention correlates to behavior, the correlation is weak. The outtake is therefore that a high purchase intention score is no guarantee of market success.

Purchase Intention is a poor predictor of behavior – so what?

So, let’s take this example a little further.

If we profile “potential users” of a new concept based on the “definitely likely to try” purchase intention measure, the profile will be based on more people who “won’t buy” the product than those who will. If we add “probably likely to try”, nearly three quarters of the people used for profiling, will not actually buy the product. This has major implications for concept optimization and targeting decisions.

The intention / behavior disconnect is impacted by the question itself, category, involvement, complexity, uniqueness and cultural response patterns to name a few contributors. And my results therefore reflect common practice in Australia and New Zealand in the FMCG category. But broadly speaking, my experience across the globe is that purchase intention is a poor predictor of future behavior irrespective of the above factors. Yet despite the shortcomings, purchase intention remains the central measure for the majority of concept tests being completed.

Weighted Purchase Intention adjusts for “average error”

There is an argument that if we weight purchase intention, we can adjust for the “average error” that is observed in studies like mine. Adjusting for average error is an admission by users that purchase intention is a poor predictor of behavior. Adjusting for error often happens at an aggregated level which means it is of little use when profiling. What we need, is a way to understand how individual people come to the decision to try a new brand.

Behavior change not beauty pageant

Most concept testing uses the 1980’s A-I-D-A approach. Stimulus simulates Awareness. The description stimulates Interest. The purchase intention measure represents Desire, with an Action the result of desire.

More current behavior change models obviously take far more factors into consideration, like context (competitive set) for example. There is a lot of good thinking out there about how people switch brands, change behavior and try new concepts. And this is where it gets interesting. When we start treating concept testing, less as a beauty pageant and more about behavior change, the more opportunity we have to understand why people try new things. If we understand why people try new things we can better predict how they will behave when faced with a choice.

Sometimes doing nothing is just easier

Despite its inherent inadequacies, purchase intention stubbornly hangs on.

We can show customers new ideas/concepts/products and ask them to change. However, some behaviors are deeply ingrained, even in the face of a superior offer. The same inertia that stops a consumer from exploring product alternatives impacts research companies and research users when evaluating the purchase intention measure. “We’ve invested in normative data focused on purchase intention so we can’t question that”.

Using purchase intention for predicting success in concept screening: A PhD

For the record, my PhD studies look at using purchase intention to predict success in concept screening. More specifically it looks at alternatives and additions to this measure (including context) improve behavioral predictions. This yields prediction success of 65% when applying more appropriate models. In future posts I hope to expand on some of my findings. Please feel free to contact me at www.linkedin.com/in/paul-epplett-catalyst-int or visit my website https://catalyst-int.com

Paul Epplett is a career market researcher, currently completing a PhD focusing on predicting success in concept screening.