Testing audience malleability

Let’s suppose you want to find out how entrenched a given audience is when it comes to a particular behaviour or attitude (regardless of whether they are doing the most efficacious thing or fully aware of all the facts when forming their opinion).

Icaro’s malleability question was developed with this in mind, with the express intention of not only testing how dug in their heels were, but also the relative effectiveness of different messages or stimuli at changing their minds.

The question is based on the respondent providing an initial score on a sliding scale from 0-100. This could be, for instance, how likely they are to do activity X in the next week or how serious an issue they think Y is (how you label the scale is flexible). It would be perfectly valid to question the use of a 100 point scale as an overly diffuse range as, for example, one person’s 60 is another person’s 75. However, the power of this question is not in the interpretation of the specific starting score per se, but in the movement each message or piece of stimuli generates from this given starting point.

After seeing each piece of information, the respondent is asked whether this makes them change their position on the scale and to move the slider accordingly. In this way, we can see the average ‘movement’ created by each one and the likely number of messages required to change views by a certain degree.

In the diagram below we’ve removed what statements A to E actually said (the example is for a client and is unpublished) but you can clearly see that the majority of the shift in opinion is achieved by looking at two statements – after that, there is a slight downward trend but it has largely plateaued. Statements A and B are the most effective, achieving a 10 point swing when they are seen first.

This shows that a) the public can be made to fairly significantly soften their stance on this given issue and b) that it only really takes one or two well-chosen statements to effect this change. The malleability construct can be applied in a number of different ways, although a central requirement is a fairly sizeable quantitative sample (e.g. 2000+), to enable the order of the messages to be randomised effectively. Please do get in touch if you think this type of question might help you, we’d love to talk about the various options.