The document discusses viewability measurement and compares metrics between two ad exchanges and advertising campaigns. Key findings include: - Direct in-ad measurement is necessary for accurate viewability reporting. On-site measurements are also valid. - Viewability results are generally consistent for campaigns running on the same demand-side platform (DSP) versus across different DSPs. - Inventory with CPMs of $0.01 performed similarly to higher-priced inventory in terms of viewability metrics on the same exchange.

1. January 2019 / Page 0marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
State of Viewability
January 2019
January 2019
Augustine Fou, PhD.
acfou [at] mktsci.com
212. 203 .7239

2. January 2019 / Page 1marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Viewability Measurement Notes
Direct measurement is necessary, measurement done in-ad
• Viewability must be measured in-ad. On-site measurements are also
valid, and this can be corroborated by simply viewing the page layout –
where the ad slots are arrayed on the page.
• Desktop and mobile web are similar since browsers are standard; but app
impressions must be separated out because measurement reliability in-
app is low; also scrolling in mobile means actual viewability is very low
• From the data, there was no significant difference whether bots were
excluded or not; this is likely because the campaigns being measured in
this study had very little bots to begin with (good buying).
• And yeah, things should add up to be 100% (internal data consistency)
hasFocus
(is browser the
active window?)
visibilityState
(is browser minimized
or tab active?)
intersection
(is the ad slot in
the viewport?)
viewable = xx

3. January 2019 / Page 2marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Exchange A vs Exchange B
Exchange A has more confirmed humans, higher quality than B
Exchange A, All CPMs exclude app app only blended
hasFocus:1 4.2% 5% 0.5% 4.8% 4.7% 5%
-hasFocus:1 84.7% 95% 10.6% 95.2% 95.3% 95%
88.9% 11.1% 100.0%
visibilityState:visible 69.9% 79% 9.6% 86.1% 79.5% 80%
-visibilityState:visible 19.0% 21% 1.5% 13.9% 20.5% 20%
88.9% 11.1% 100.0%
intersection >50% 20.1% 23% 4.2% 37.4% 24.2% 24%
-intersection >50% 68.8% 77% 7.0% 62.6% 75.8% 76%
88.9% 11.1% 100.0%
viewable 0.8% 1.5% 0.9%
(excl hasFocus:1) viewable 17.8% 32.2% 19.3%
Exchange B, All CPMs exclude app app only blended
hasFocus:1 0.4% 0% 0.0% 0.0% 0.4% 0%
-hasFocus:1 95.1% 100% 4.5% 100.0% 99.7% 100%
95.5% 4.5% 100.0%
visibilityState:visible 34.3% 36% 4.2% 92.2% 38.4% 38%
-visibilityState:visible 61.2% 64% 0.4% 7.8% 61.6% 62%
95.5% 4.5% 100.0%
intersection >50% 8.0% 8% 0.5% 10.9% 8.5% 9%
-intersection >50% 87.5% 92% 4.0% 89.1% 91.5% 91%
95.5% 4.5% 100.0%
viewable 0.0% 0.0% 0.0%
(excl hasFocus:1) viewable 3.0% 10.0% 3.3%

4. January 2019 / Page 3marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
2 Campaigns on the same DSP
Quality and viewability are consistent for campaigns on same DSP
Campaign A, All CPMs exclude app app only blended
hasFocus:1 1.1% 1% 0.7% 4.8% 1.8% 2%
-hasFocus:1 83.6% 99% 14.6% 95.2% 98.2% 98%
84.7% 15.3% 100.0%
visibilityState:visible 61.1% 72% 13.7% 89.3% 74.8% 75%
-visibilityState:visible 23.6% 28% 1.6% 10.7% 25.2% 25%
84.7% 15.3% 100.0%
intersection >50% 17.1% 20% 7.4% 48.2% 24.5% 24%
-intersection >50% 67.6% 80% 7.9% 51.8% 75.5% 76%
84.7% 15.3% 100.0%
viewable 0.2% 2.1% 0.3%
(excl hasFocus:1) viewable 14.6% 43.1% 18.3%
Campaign B, All CPMs exclude app app only blended
hasFocus:1 1.0% 1% 0.5% 2.7% 1.5% 2%
-hasFocus:1 80.8% 99% 17.7% 97.3% 98.5% 99%
81.8% 18.2% 100.0%
visibilityState:visible 57.8% 71% 12.9% 70.5% 70.6% 71%
-visibilityState:visible 24.0% 29% 5.4% 29.5% 29.4% 29%
81.8% 18.2% 100.0%
intersection >50% 11.2% 14% 5.2% 28.5% 16.4% 16%
-intersection >50% 70.6% 86% 13.0% 71.5% 83.6% 84%
81.8% 18.2% 100.0%
viewable 0.1% 0.5% 0.2%
(excl hasFocus:1) viewable 9.7% 20.1% 11.6%
OBSERVATIONS
• Quality is virtually
identical (dark blue and
dark red similar across
both campaigns)
• Buying on the same DSP
appears to yield very
similar viewability results
across campaigns
• % of impressions from
apps vs non-apps are
consistent across
campaigns

5. January 2019 / Page 4marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
1cent CPM vs All CPMs
No statistical difference between 1cent CPM inventory vs All CPMs
Exchange A, 1c CPM exclude app app only blended
hasFocus:1 4.9% 5% 0.0% 0% 4.9% 5%
-hasFocus:1 92.0% 95% 3.1% 100% 95.1% 95%
96.9% 3.1% 100.0%
visibilityState:visible 57.3% 59% 1.8% 59% 59.1% 59%
-visibilityState:visible 39.6% 41% 1.3% 41% 40.9% 41%
96.9% 3.1% 100.0%
intersection >50% 30.8% 32% 1.5% 48% 32.3% 32%
-intersection >50% 66.1% 68% 1.6% 52% 67.7% 68%
96.9% 3.1% 100.0%
viewable 0.9% 0.0% 0.9%
(excl hasFocus:1) viewable 18.8% 28.4% 19.1%
Exchange A, All CPMs exclude app app only blended
hasFocus:1 4.2% 5% 0.5% 4.8% 4.7% 5%
-hasFocus:1 84.7% 95% 10.6% 95.2% 95.3% 95%
88.9% 11.1% 100.0%
visibilityState:visible 69.9% 79% 9.6% 86.1% 79.5% 80%
-visibilityState:visible 19.0% 21% 1.5% 13.9% 20.5% 20%
88.9% 11.1% 100.0%
intersection >50% 20.1% 23% 4.2% 37.4% 24.2% 24%
-intersection >50% 68.8% 77% 7.0% 62.6% 75.8% 76%
88.9% 11.1% 100.0%
viewable 0.8% 1.5% 0.9%
(excl hasFocus:1) viewable 17.8% 32.2% 19.3%
OBSERVATIONS
• Quality is virtually identical
(dark blue and dark red similar
across both CPM ranges)
• Buying on the same exchange
appears to yield very similar
viewability whether the CPMs
were 1cent or all CPMs
• % of impressions from apps is
higher for higher average
CPMs

6. January 2019 / Page 5marketing.scienceconsulting group, inc.
linkedin.com/in/augustinefou
Viewability vs Ad Blocking
Measurement accuracy depends on a number of factors
Viewability Ad Blocking
• Is dependent on the page
layout, where the ad is
• Must be measured in-ad;
direct measurement is key
• In-app impressions must be
excluded; desktop and
mobile we are similar
• Is related to the browser/
tech used by the user
• Must be measured on-site;
direct measurement is key
• Desktop must be separated
from mobile
• Bots should be excluded, %
of data not measurable
should be disclosed