Research: behavioral KPI VS branding KPI correlation
1. Can Rich Media Metrics
Predict Brand Impact?
Ken Mallon Rick Bruner
SVP, Custom Solutions Head of Research, NA Sales
Dynamic Logic Google
2. Situation Overview
Ad serving metrics (click rate, rich media
interactions) are standard direct-response measures
of campaign performance
Some advertisers care more about brand objectives
than direct response online
Third-party survey-based test/control experiments
have become the norm for measuring lift in brand
attitudes
3. Key Research Question
Could ad server metrics (click rate, rich media
interactions, expansions) be proxies for assessing
brand performance of campaigns?
4. Methods (Database Construction)
Rich media ads served by DoubleClick that were also
measured by Dynamic Logic
Final analysis dataset
– blinded to advertiser
– 4,299 records (creative units)
– Contained both ad interaction data and brand impact metrics
Merging was performing by independent 3rd party to
maintain the blind
5. Methods (Metric Definitions)
Behavioral metrics
– Interaction
A person is said to interact with a rich media ad if they hover over it for at least
one second
– Click-through-rate
– Expansion rate
Percent of impressions in expandable format that generate an expansion
Brand metrics
– Aided brand awareness
– online ad awareness
– message association
– brand favorability
– purchase intent
6. Methods (Statistical Analyses)
Correlation
– Each brand metric was correlated with each behavioral metric
– This was done both on the original scale as well as log-
transformed scale for behavioral metrics
Linear regression
– Models were developed to predict the brand impact metrics as
a function of each of the log-transformed behavioral metrics
– Adjustment variables included baseline brand levels (brand
levels within each ad campaign among those not exposed to the
ads), the category of the advertised brand and other factors
8. Results
Relationship between Interaction Rates and Brand Metrics
Weak positive relationship between ad interaction rates and both ad awareness
and brand favorability
Negative relationship with message association
Regression analyses revealed statistically significant but practically unimportant
relationships (r-squares in the range of 1-3% for the 4 models)
0.15 0.132
0.120
0.10
0.05
0.00
Ad Awareness Message Brand Opinion Intent
Association
-0.05
-0.048
-0.10
-0.15
-0.162
-0.20
9. Results
Relationship between Expansion Rates and Brand Metrics
Weak positive relationship between ad interaction rates and ad awareness
Negative relationship with purchase intent
Regression analyses revealed statistically significant but practically unimportant
relationships (r-squares in the range of 2-4% for the 4 models)
0.25
0.20 0.188
0.15
0.10 0.071
0.05
0.00
Ad Awareness Message Brand Opinion Intent
-0.05 Association
-0.053
-0.10
-0.15
-0.20 -0.190
-0.25
10. Conclusions
Ad behaviors not good predictors of brand impact
These results show that rich media ad behaviors such
as clicking, interaction and expanding are not good
predictors of the branding impact of ads
It may be that people interact with ads that are eye-
catching or have an interest game, etc. but that this
activities may actual distract from delivering brand
messages and other brand attributes
Suggest using copy-testing as a better predictor of in-
market branding success