Evolution of tracking incrementality from PSA to continuous uplift tracking. Why retargeting in Gaming is challenging. Calculating retargeting effects. How to approach mobile retargeting?
2. What We Do: App Retargeting
Existing
User
They open
your app…
But leave without
purchasing
Later, they‘re
browsing another
app
Your ad brings them
back
Your user
converts
3. Retargeting: Shopping was easy
Online Shopping
● Easy to decide what to show
● Short-term interest
● Product is common
● Easy-to-understand value proposition
● High direct reward if successful
4. Retargeting for Gaming is
harder
Mobile Gaming
● In most cases, user doesn‘t want to buy the
product
● Mid- to long term interest
● Unique product proposition
● value proposition is hard to understand
● No direct reward
5. Two Different Challenges
Shopping
• Use dynamic templates
• Show users what they want
• Retarget them immediately
Gaming
● What to show users?
● What do they want?
● Should I wait?
6. Retargeting with a leading European Game Developer
Segmentation
Types of Creatives
● Previous Purchasers
● Inactive for X days
Generic Ad Seasonal Ad (high relevance)
9. Uplift Test Results
C O N T R O L G R O U P
R E V E N U E
TEST GROUP
REVENUE
Revenueduetoother
marketingactivities
Incremental Revenue
Organicrevenue
TEST GROUP
REVENUE
C O N T R O L
G R O U P
R E V E N U E
= –
(achievedwith Remerge) (achievedwithout)
How could we know that?
16. Client Double-Checked
Control
● Is the control group representative?
● Is the number of “whales“ approximately the same
between the test and control?
● Is the traffic comparable?
Test
18. After weeks
30-60%Uplift CR
Inc. Revenue GREAT
Inc. ROAS 120-220%
Results: PSA Test
Conversion Rate %
Test Group
Control Group
ButwhyI shouldpayfor the
control groupads?
Canwekeepit running?
24. Total Audience
Test 75% Control 25%
Won,
Impression
Lost,
No Impression
May Win May Lose
Total Reachable Audience
25. Total Audience
Test 75% Control 25%
Won,
Impression
Lost,
No Impression
May Win May Lose
That would
be a valid
comparison
Total Reachable Audience
26. Total Audience
Test 75% Control 25%
Won,
Impression
Lost,
No Impression
May Win May Lose
But we don’t know that
Total Reachable Audience
27. Total Audience
Test 75% Control 25%
That’s why we calculate uplift based
on perfectly randomized groups to
avoid bias
Total Reachable Audience
28. Methodology Check
Regular PSA Continuous Uplift Tracking
Frequency One-off Test Always On
Cost Advertiser pays PSA No extra cost
Reporting One-off Report Regular Reporting
Optimization No Optimization Continuous Optimization for
incrementality
30. • Helps to measure
important business
metrics
• Calculate the
influence of
marketing activities
on revenue
• Incrementality is
important for UA as well
• It’s much easier to
calculate in retargeting
• Different
methodologies
• Required a lot of data
to achieve statistical
significance
Why Should I Care
31. Thank you! Have questions for us?
Give Aleks a call
+491754874914
Or send us an email
aleksandr@remerge.io