Mike Follett, MD, Lumen Research MAD//Fest London presentation on what eye tracking technology tells us about advertising engagement across media and platforms.

1. Paying (for) attention
How eye tracking reveals the true nature of attention
Mike Follett
Lumen Research
November 2018

2. Attention technology
An attention currencyfor the attention economy
Eyetrackingand attention predictions at speed and scale for desktop and mobile
• Cross media
• Programmatic integrations
• Link tosales

3. Methodology
Continuous panel
n=500, UK
2016-present
Bespoke studies
n=21,600 (UK, US, Aus, SE, CH, FR,
NL)
2013-present

4. Source: Lumen Research
Basedon attentiondata from the Lumen AttentionPanel, Jan 2016-present
Viewable = 50% of pixelsfor 1+sec
Examplefullpage studyheatmap
Theattention funnel: a big difference between viewability and viewing

5. Source: Lumen Research
Basedon attentiondata from the Lumen AttentionPanel, Jan 2016-present
Viewable = 50% of pixelsfor 1+sec
173k impressions served
59% viewable
20.2 seconds av. viewable
time
20% viewed the ad
1.3 seconds av
view time
3% viewed for1
sec+
Examplefullpage studyheatmap
Theattention funnel: a big difference between viewability and viewing

6. Dramatic differences in the quality of attention across different media and platforms
Source: Lumen Research
Based on attention data from the Lumen Attention Panel, Jan 2016-present + bespoke studies
Viewable = 50% of pixels for 1+ sec
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
%ofviewableimpressionsseenformorethan
Dwell time (s)
Desktop pre-roll
OOH
Press
Mobile Display
Facebook Desktop
Desktop Display

7. Key drivers of differences in quality of attention
0
0.5
1
1.5
2
2.5
3
3.5
4
0 10 20 30 40 50 60
Averagedwelltime(s)
Viewable time (s)
All Digital ads
All print ads

8. Key drivers of differences in quality of attention
0%
10%
20%
30%
40%
50%
60%
70%
0 20 40 60
%Viewed
Viewable time (s)
0
0.5
1
1.5
2
2.5
3
3.5
4
0 10 20 30 40 50 60
Averagedwelltime(s)
Viewable time (s)
All Digital ads
All print ads

9. Key drivers of differences in quality of attention
0%
10%
20%
30%
40%
50%
60%
%viewed
100 top ranked domains
0%
10%
20%
30%
40%
50%
60%
70%
0 20 40 60
%Viewed
Viewable time (s)
% likelihood
to view
MPUs
0
0.5
1
1.5
2
2.5
3
3.5
4
0 10 20 30 40 50 60
Averagedwelltime(s)
Viewable time (s)
All Digital ads
All print ads

10. Non-Skippable 6s, n = 169
Non-Skippable 15s, n = 32
Non-Skippable 20s, n = 29
Skippable, n = 769
Pre-roll performance driven by non-skippable ads
0%
20%
40%
60%
80%
100%
0 5 10 15 20 25 30
%stillwatching(inc.non-viewableimpressions)
Seconds into ad
6`` Non-Skippable 15`` Non-Skippable 20`` Non-Skippable Skippable

11. Non-Skippable 6s, n = 169
Non-Skippable 15s, n = 32
Non-Skippable 20s, n = 29
Skippable, n = 769
Aggregated heatmap of % viewed to skippable pre-roll
Aggregated heatmap of Av dwell time
Pre-roll performance driven by non-skippable ads
0%
20%
40%
60%
80%
100%
0 5 10 15 20 25 30
%stillwatching(inc.non-viewableimpressions)
Seconds into ad
6`` Non-Skippable 15`` Non-Skippable 20`` Non-Skippable Skippable

12. Facebook: strong results driven by ‘in line’ layout
Source: Lumen passive desktop panel. Display (173k impressions – MPU, Skyscraper, Leaderboard, Billboard, DMPU), Facebook In-Feed (23k)
59%
41%
Display Average Facebook In-Feed
% Viewable (50%, 1")
20%
30%
Display Average Facebook In-Feed
% of viewable ads seen
1.3
2.1
Display Average Facebook In-Feed
Average dwell time per view
(s)
Example heatmap from user browsing session
(6.30am to 12.30am)

13. Why this matters: Attention leads to recall…
 Source: Lumen lab studies; 57 digital ads n=400
8%
12%
16%
18%
17%
22%
0%
5%
10%
15%
20%
25%
Not seen 0 to 1s 1 to 2s 2 to 3s 3 to 4s 4 to 5s
Promptedrecall
Dwell time on advertising

14. …and to sales
47
86
97
154 158
168
0 to 10 10 to 20 20 to 30 30 to 40 40 to 50 50 to 60
Conversionrate(Indexed
camaignaverage=100
Predicted % viewed
82 73
100
131
175
216
0 to 0.5 0.5 to 1 1 to 1.5 1.5 to 2 2 to 2.5 2.5 to 3
Conversionrate(Indexed
camaignaverage=100
Predicted average dwell time (s)

15. …and to sales
Recent project redirected media
budget towards ‘high attention’ sites
and formats in a test region
239%increase in sales conversions vs
control region
ROI = 10x
47
86
97
154 158
168
0 to 10 10 to 20 20 to 30 30 to 40 40 to 50 50 to 60
Conversionrate(Indexed
camaignaverage=100
Predicted % viewed
82 73
100
131
175
216
0 to 0.5 0.5 to 1 1 to 1.5 1.5 to 2 2 to 2.5 2.5 to 3
Conversionrate(Indexed
camaignaverage=100
Predicted average dwell time (s)

16. Creative matters too
100
180
Without Penguin With Penguin
% viewing (index)

17. Using attention data to price inventory across media
2.5 4.1 5.2
12.6
24.8
127.4
Desktop display Facebook (desktop) Mobile display DOOH (2m head-on) Press YouTube (Desktop)
Dwelltimeminutesper000impressions/contacts/readers
Attention (min) per 000 impressions
Source: Lumen Research
Based on attention data from the Lumen Attention Panel, Jan 2016-present + bespoke studies
Viewable = 50% of pixels for 1+ sec
Desktop pre-roll
Assumes 100% viewability for press

18. These attention predictions can be used create ‘attention segments’ to inform
programmatic buying
Attention model
- Viewable ms
- Size
- Domain
- Day of the week
Viewability data
- Placement ids
- Size
- Viewable ms
AAA: 200% + effective
attention
AA: 150% - 200%
effective attention
A: 100% - 150%
effective attention
B: 50% - 100%
effective attention
C: <50% effective
attention
Attention segments
Data fed and processed once an hour, meaning segments find placements currently receiving most attention
e.g.

19. Implications for advertisers, planners and publishers
All impressions are not created equal
Buy (and sell) attention, not impressions
Optimise creative AND media for attention

20. Contact
Mike Follett
mike.follett@lumen-research.com
Managing Director
Fiona Evans
Fiona.evans@lumen-research.com
Sales Director
22 Highbury Grove
Unit 215, ScreenWorks
London, N5 2EF
020 3735 5199
www.lumen-research.com