🧠 How to Apply Psychology and Data to your Influencer Activity
How Influencers Overcome Our Psychological Defences
One of the biggest barriers to nudging is ‘source credibility’. When people suspect they are being influenced, they tend to put their conscious defences up and become more critical of the message, making it less persuasive.
You know, for example, that a skin cream advert is trying to persuade you to buy skin cream, so when it says ‘eight out of ten people noticed visible signs of youthinisation thanks to hyalubolloxin acid* (*survey of five people)’, you’ll probably be sceptical of the message and less influenced by it.
When people suspect they are being influenced, they tend to put their conscious defences up and become more critical of the message, making it less persuasive. (Photo: Monstera / Pexels)
This is where ‘meta-nudging’ comes in. This approach changes behaviour through the power of social influencers, who spread nudges to large groups of people from the top-down, like a pyramid. As one economics paper put it,
“…one can also successfully nudge individuals indirectly by harnessing the power of social norms enforcement. That is, by targeting those who enforce behavior – rather than those whose behavior one wants to alter – behavioral interventions would aim at nudging individuals in positions of power who have the ability to enforce the transgressors’ adherence to social norms.”
This is the power of social influencers. They have a wide reach, they set social norms, and – since they are trusted, familiar and unrelated to nudgers – bypass source credibility defences.
Social influencers can bypass source credibility defences. (Photo: Ivan Samkov / Pexels)
However, there is a science to it.
For example, influencer posts get more engagement (18% higher on TikTok) when they use concrete language (e.g., ‘These slippers are soft’ rather than ‘These slippers are great’).
There are a few frameworks which explain how to make influencer content go viral (like Made to Stick, Contagious, and our own Hooked: Why cute sells and other marketing magic we can’t resist), but they have the same common themes. The content should be emotional, surprising, concrete (e.g., visual or visualisable), simple, curious, and relevant to the target audience.
But, we know that behavioural science is not one-size-fits-all. (More on this topic here)
The nudges that work with a crypto influencer, for instance, will be quite different to those that work with a cookery influencer. Fortunately, data science can help here.
By applying feature extraction, clustering, and personality predictive models, we can automatically recommend the influencers and the messaging elements that will work best for a given target audience.
1 - Start with a Seed
Here’s how it works. Firstly, you give us a ‘seed’ of a particular topic you are interested in. It might be a hashtag for a product, an account’s followers, or a track used in TikToks, for instance.
2 - Feature Extraction
We then automatically scrape the data (text, images, audio, and video, as well as metadata such as ‘likes’ and shares) associated with that seed and extract the features from it. With video, for example, these features might include the presence and number of faces, the facial features, the emotional expressions, the colours, the rate of movement, and so on.
3 - Content Clustering
Next, we conduct a clustering analysis to see how the content clusters into segments (for example, football videos with lots of green and movement versus recipe videos with lots of brown and overlaid text).
4 - Predictive Modelling
The metadata tells us which content cluster has the most viral potential, while predictive models applied to the clusters tell us what kind of audience it would work on and why (for example, videos with bright colours are liked by extraverts). These models can predict what kind of influencers to approach for appropriate viral content (like upbeat celebrities for those extraverts) and what kind of brief to give them (be social, loud and funny for those extraverts).
5 - Does it work?
With this approach, we reduced CPM by 8% for one client, while saving them thousands of pounds and weeks of wasted time.