Using predictive models of attention to link attention to sales
Bill Bernbach said: “You can’t sell a man who isn’t listening.” Or looking, we might add. But is that true? Is unseen unsold?
To assess the relationship between attention and sales, we can use Lumen’s predictive models of attention to understand the attention digital advertising is likely to generate and how this relates to online sales.
1
Using data from Lumen’s desktop and mobile panels in the UK and now the US, we can build a model that predicts how likely someone is to actually view an ad when it appears in a given format, in a given location on a given website, and for a given amount of time.
2
We can then apply these predictions to impression-level campaign data for digital campaigns. This gives us an estimate of how much actual attention each impression in a campaign is likely to have generated.
3
Finally, we can look at the relationship between the ads that we predict get lots of attention and the ads that actually generated lots of sales. If attention leads to sales, then the ads that are most likely to be noticed – or most likely to gain a lot of viewing time – should also be the ones that lead to the most sales.
It should be noted that these predictions are just about media characteristics – they don’t take into account creative or targeting. The basic model assumes averagely attention-grabbing creative served to averagely well-targeted consumers, though it is possible to add these factors into client-specific models. Given what we know about how people look at advertising for ‘their’ brands or ads that are particularly relevant for them, this is probably an advantage. Holding creative and targeting constant allows us to isolate the impact of media factors on attention and sales. Working with British Gas and Mediacom in the UK, we have been able to use the Lumen prediction model to calculate the likely attention British Gas’ digital advertising is generating and how this relates to sales. The results are reassuring. There is a strong relationship between attention and sales. Ads that, while technically viewable, have a very low chance of getting looked at don’t tend to result in sales. Ads that have a higher chance of being seen seem to have a higher chance of converting. Similarly, we’ve found that the ads we predicted gained the longest view time were also the most likely to generate online sales for British Gas. The natural next step for British Gas was to use this insight not only to understand the past but inform the future. So, we created a bidding strategy for them to be able to buy high-attention inventory and set up a test to see whether buying ads that people are more likely to look at has an impact on sales. In the region where British Gas bought high-attention inventory, post-click conversions increased by 309% while post-view conversions increased by 200%, all other factors being held constant. Since these initial tests, Lumen has conducted many similar studies and for a variety of advertisers in different categories. The results are always the same: more attention equals more sales, though the strength of the relationship differs from category to category and brand to brand. Attention leads to sales, though how much attention and how many sales depends on media planning, creative design, and audience targeting. It seems that maximising your attention currency can enhance return on investment.
