Media buying

+ Attention stack


Buy attention, not impressions

The Lumen Attention Stack is a contextual targeting tool that enables advertisers to evaluate digital ad inventory on the basis of the visual attention it receives.



Eye tracking panel

How the eye tracking works

Lumen’s software collects and combines two different types of data:
  • Browser extension records events on device
  • Webcam eye tracking records visual attention via front-facing camera

Panel make up

UK: 250 desktop and 750 mobile users; nationally representative sample
US: 1000 mobile users nationally representative sample
Panels refreshed frequently

Data security

All panellists fully informed, consented and compensated for participation.
No personally identifiable information (PII) collected, stored or reported.
All data collection, storage and processing fully GDPR-compliant

Attention prediction model

Combining placement and attention data

Data from the Lumen panel tells us
Placement data: What ads could be seen (what was viewable, in what location and for how long, etc.)
Attention data: What ads people did, in fact, see, and for how long

Predicting attention from placement data

Placement and attention data from the eye tracking panel is used to power a predictive model of attention
Estimates likely levels of attention based on a number of placement characteristics
Attention to ad predicted every 100ms of exposure

Placement data included in model

Ad format, % pixels in view, Viewable time, Screen real estate, Scroll speed, Number of competing ads on screen, ad positioning and page geometry, Domain, Etc.



Lumen collects attention data from our proprietary eye tracking panels


We create a predictive model of attention that calculates the ‘attentive seconds’ advertising receives, based on publisher, format and viewability metrics.


Attentive second predictions incorporated within your DSP or Google Custom Algorithm to allow you to buy ‘high attention’ inventory (and avoid ‘low attention’ inventory) in real time


Attention reports linked to brand lift studies and sales outcomes

What is an attentive second?

Our attention research has shown that people are very good at ignoring advertising. Just because an ad is 100% viewable doesn’t mean that it will be 100% viewed. And just because an ad was viewable for 10 seconds does not mean that it will actually viewed for the full 10 seconds.

We use eye tracking to understand the reality of attention: the attentive seconds generated by an ad per 1000 impressions.
It works like this.

Imagine you buy 1000 desktop MPU impressions.
Some of them are non-viewable. Imagine that 57% of the ads are viewable.

Of the ones that are viewable, the average viewable time might be 18 seconds.

But that doesn’t mean that 570 ads have been seen for 18 seconds.
In fact, only about 17% of viewable impressions, or 8% of total impressions actually get viewed.

And of those, the average attention time that the ads receive is around 1.3 seconds.

So, when you look at the actual amount of attention 1000 ads receive, you have to understand how many people actually looked, and how long they actually looked for.

In this case, 8% of impressions x 1.3 seconds x 1000 impressions = 110 attentive seconds per 100 impressions


British Gas


+ Brand-specific attention stacks