Professional Documents
Culture Documents
xx
Your customers
attributed to yourinfluenced
ad efforts by
through
your ad
click-based
efforts measurement
Source: Facebook Internal Data, Lift Studies, Feb 2017. Based on 2,177 Facebook Lift studies with online sales-based objectives.
The consumer path is complex
Measurement with
cookies alone tells a
partial story
Facebookresearch solidifies
the power of incrementality
measurement
23Average disagreement in
% Incrementality
measurement disagrees with
one-day click attribution
winning tactic chosen
between the two systems systems on which
marketing tactics are most
effective.
Source: Facebook Analysis of 580 Conversion Lift A/B tests run Feb 2015 – Sep 2017. Percentages quoted represent aggregate overall average among all tests analyzed (statistically significant
result @ 90% CI); Oct 2017.
48%
Difference in winning tactic
is more pronounced with
website purchases
Source: Facebook Analysis of 580 Conversion Lift A/B tests run Feb 2015 – Sep 2017. Percentages quoted represent aggregate overall average among all tests analyzed (statistically significant
result @ 90% CI); Oct 2017.
The two systems disagree across various outcomes, including
mobile, with advertisers missing out on potential CPA
improvements
Rate found at which click attribution winner differs Difference in CPA when click attribution and
from incrementality incrementality disagree on winner
Website Website
Purchases 48% Purchases 79%
Source: Facebook Analysis of 580 Conversion Lift A/B tests run Feb 2015 – Sep 2017. Percentages quoted represent aggregate overall average among all tests analyzed (statistically significant
result @ 90% CI); Oct 2017.
Four steps to applying experimentation and learning to your
marketing organization
2
20% lift
TEST TEST
AD IMPRESSION
CONTRO CONTRO
L NO AD L
IMPRESSION
1. 2. 3. 4. 5.
Identify the audience Facebook randomizes Ads are delivered to You pass your Facebook compares
you want to target and your audience, then the test group conversion data to the conversions in the
the business objective separates people into Facebook via the test and control groups
you want to measure test and control groups Facebook pixel(s), a to calculate lift, then
secure data upload or posts results in Ads
app events Manager
Note: Lift studies use randomized test and control groups, with some people who will not see ads for your campaign. Testing can sometimes result in a smaller-sized test
12
audience with comparatively higher CPMs.
Your
results in
Ads Manager
• Prove ROAS
• Understand which channels drive
impact Control No Ad Impression
1 2 3 4
Facebook and client Determinethe source Run the lift test Facebook and client
determine the business of conversion data analyze lift results
objective and key Set
together
up Facebook
campaign KPIs
pixels, app events or Agree upon next
offline data source steps for attribution
D ESIRED OUTCOME AND C ONFIGURE THE R UN THE LIFT TEST A NALYZE RESULTS
KPI’S PIXEL, SDK, OFFLINE
N EXT STEPS
CONVERSIONS
15
Thank you
Appendix
Legend:
Person
Facebook ad
This came from the
did not enter
exposed group (i.e.
viewable area
reached)
Reached:
Saw Facebook
ad Lift % = ?
Facebook ad
Incremental Converters did not enter
viewable
Baseline Converters area
Reached:
Clicked
Facebook ad
This is the number of
converters in the reached
group had there not been a
Facebook ad
*Scaled to be the same size as the test group Data is for illustrative purposes only. Not actual results.
Legend:
Person
Reached:
Saw Facebook
ad Lift % = ?
Facebook ad
( 5656 17
17 ) did not enter
viewable
Baseline Converters area
Reached:
Clicked
Facebook ad
This is the number of
converters in the reached
group had there not been a
Facebook ad
*Scaled to be the same size as the test group Data is for illustrative purposes only. Not actual results.
Legend:
Person
Denominator is defined as ‘number of converters in the reached group had there not been a Facebook
ad’
100 people in test group 100 people in control group*
(opportunity to see Facebook + other ads) (opportunity to see other ads only)
Reached:
Saw Facebook
ad Lift % = ?
Facebook ad
( 5656 17
17 ) did not enter
Baseline Converters: Converters in viewable
area
Baseline Converters
this group had there not been a
Reached: Facebook ad
Clicked
Facebook ad
This is the number of
converters in the reached
group had there not been a
Facebook ad
*Scaled to be the same size as the test group Data is for illustrative purposes only. Not actual results.
Legend:
Person
vs. the hypothetical number in that group had there not been an ad
100 people in test group 100 people in control group*
(opportunity to see Facebook + other ads) (opportunity to see other ads only)
Reached:
Saw Facebook Lift % = 300%
ad
Facebook ad
( 5656 1717 ) did not enter
Baseline Converters: Converters in viewable
Reached: Facebook ad
Clicked
Facebook ad
This is the number of
converters in the reached
group had there not been
a Facebook ad
*Scaled to be the same size as the test group Data is for illustrative purposes only. Not actual results.