Noah Freeman, CTO & Co-Founder
Social Fulcrum
July 23, 2019
SESSION:
Incremental Lift Testing: Determine the
True Impact of Your Ad Campaigns
Deep Dive
2
Lift Testing: An Example
What % is real? True ROI
30% 3x
100% 4x
50% 3.5x
Audience In-Platform ROI
Cart abandoners 10x
New customers 4x
Total 7x
True ROI is above target,
so increase total budgets
Raw data tells us to move
money to cart abandoners....
...but real data says
the opposite!
Imagine knowing the real impact of each ad...
3
$
$
Who Can Lift Test
4
5
Lift testing: available for online & in-store purchases
Purchase location Online In-store
Time period Raw ROI Lift-adjusted ROI Raw ROI Lift-adjusted ROI
Pre-2015 ✓
2015-2016 ✓ ✓ ✓
2017-now ✓ ✓ ✓ ✓
Why We Lift Test
6
Attribution tells you if an ad preceded a purchase,
NOT if it caused the purchase
7
Saw ad
Made purchase
Did the pixel
give credit?
Did the ad cause
the purchase?
Monday 9am
Tuesday 8am
Yes
Yes!
Monday 9am
No purchase
No
No!
Monday 9am
8 days later
No
Yes - but 8 days
later
Monday 9am
Tuesday 8am
Yes
No - would have
bought anyway
Lift Based Attribution
Time Based
Attribution
Lift testing tells us what %
of the reported purchases
are caused by the ad
8
Lift testing lets us give correct, partial credit to purchases
across a huge time window
How We Lift Test
9
20%
Control
StartingAudience
See real
ad
10 purchases per
1K audience
80%
Exposure
Incremental Lift =
7 Purchases
See
charity
ad
3 purchases per
1K audience
■ Before: 100% of credit for
7 days of purchases
■ After: 70% of credit for 90
days of purchases
■ Outcome: More accurate,
higher ROI, more scale
10
To run a lift test, you show real ads to your
exposure group, and fake ads to a control group
11
To do a great test, you need to get a few things right
03
02
05
01
04
Design the test to give you
audience-level lift results
Make sure you can measure
ALL purchases, anywhere
Use good software, or big
control groups
Make sure your control group
is dark for 1-2 months pre-test
Find the right partner
• Most systems give only “whole-test” level incrementality
• Incrementality by audience is better
• Customers wander through your funnels in 90 days
• If you can’t measure all your checkouts, data suffers
• Control groups cost money and customers
• Accurate, small control groups are only possible with great third-party
software
• Customers who switch to control ads have “purchase momentum” from
previous real ads
• This momentum overstates control group purchases, and shows incorrect
low incrementality results
• Lift tests: easy to try, hard to do well
• You need to choose the right partner
12
You can choose from many partner types
Ad platforms
Pro: cheapest choice,
simple setup
Con: perception of bias,
HUGE control groups, no
audience-level data
Example: Facebook’s
“Conversion Lift Test”
Big measure
vendors
Pro: You may already
have a contract in place
Con: Expensive, no
audience-level data, long
delays on reporting
Example: ODC
Measure
boutiques
Pro: Granular data, small
control groups
Con: None
Example: Social Fulcrum
What Results Do We Typically See?
13
14
Lift tests show incrementality between 25% and 95%,
leading to real ROIs from 3-25x
Ad type
% of reported
revenues that are
real
Prospecting 90-95%
Retargeting
(Site Traffic)
25-35%
Retention/reactivation
(Past Customers)
35-45%
Advertiser Real ROI
5-10x
low scale 5-15x
high scale 3-5x
15-25x
● Lift testing tells you the real impact of your digital ads
● Recent tests for retailers have shown tremendous ROI
● You need to choose the right partner to meet your needs and deliver
the details that are important to you
Now what?
15
Q & A
16
● Do you ever doubt the results your platforms are reporting? Why?
● How have you dealt with this before?
● Have you run a lift test? How did it improve your decision-making?
● For those who have not run a lift test, how could lift testing help your
business?
Discussion questions
17

Incremental lift testing: Determine the true impact of your ad campaigns

  • 1.
    Noah Freeman, CTO& Co-Founder Social Fulcrum July 23, 2019 SESSION: Incremental Lift Testing: Determine the True Impact of Your Ad Campaigns Deep Dive
  • 2.
  • 3.
    What % isreal? True ROI 30% 3x 100% 4x 50% 3.5x Audience In-Platform ROI Cart abandoners 10x New customers 4x Total 7x True ROI is above target, so increase total budgets Raw data tells us to move money to cart abandoners.... ...but real data says the opposite! Imagine knowing the real impact of each ad... 3 $ $
  • 4.
  • 5.
    5 Lift testing: availablefor online & in-store purchases Purchase location Online In-store Time period Raw ROI Lift-adjusted ROI Raw ROI Lift-adjusted ROI Pre-2015 ✓ 2015-2016 ✓ ✓ ✓ 2017-now ✓ ✓ ✓ ✓
  • 6.
  • 7.
    Attribution tells youif an ad preceded a purchase, NOT if it caused the purchase 7 Saw ad Made purchase Did the pixel give credit? Did the ad cause the purchase? Monday 9am Tuesday 8am Yes Yes! Monday 9am No purchase No No! Monday 9am 8 days later No Yes - but 8 days later Monday 9am Tuesday 8am Yes No - would have bought anyway
  • 8.
    Lift Based Attribution TimeBased Attribution Lift testing tells us what % of the reported purchases are caused by the ad 8 Lift testing lets us give correct, partial credit to purchases across a huge time window
  • 9.
  • 10.
    20% Control StartingAudience See real ad 10 purchasesper 1K audience 80% Exposure Incremental Lift = 7 Purchases See charity ad 3 purchases per 1K audience ■ Before: 100% of credit for 7 days of purchases ■ After: 70% of credit for 90 days of purchases ■ Outcome: More accurate, higher ROI, more scale 10 To run a lift test, you show real ads to your exposure group, and fake ads to a control group
  • 11.
    11 To do agreat test, you need to get a few things right 03 02 05 01 04 Design the test to give you audience-level lift results Make sure you can measure ALL purchases, anywhere Use good software, or big control groups Make sure your control group is dark for 1-2 months pre-test Find the right partner • Most systems give only “whole-test” level incrementality • Incrementality by audience is better • Customers wander through your funnels in 90 days • If you can’t measure all your checkouts, data suffers • Control groups cost money and customers • Accurate, small control groups are only possible with great third-party software • Customers who switch to control ads have “purchase momentum” from previous real ads • This momentum overstates control group purchases, and shows incorrect low incrementality results • Lift tests: easy to try, hard to do well • You need to choose the right partner
  • 12.
    12 You can choosefrom many partner types Ad platforms Pro: cheapest choice, simple setup Con: perception of bias, HUGE control groups, no audience-level data Example: Facebook’s “Conversion Lift Test” Big measure vendors Pro: You may already have a contract in place Con: Expensive, no audience-level data, long delays on reporting Example: ODC Measure boutiques Pro: Granular data, small control groups Con: None Example: Social Fulcrum
  • 13.
    What Results DoWe Typically See? 13
  • 14.
    14 Lift tests showincrementality between 25% and 95%, leading to real ROIs from 3-25x Ad type % of reported revenues that are real Prospecting 90-95% Retargeting (Site Traffic) 25-35% Retention/reactivation (Past Customers) 35-45% Advertiser Real ROI 5-10x low scale 5-15x high scale 3-5x 15-25x
  • 15.
    ● Lift testingtells you the real impact of your digital ads ● Recent tests for retailers have shown tremendous ROI ● You need to choose the right partner to meet your needs and deliver the details that are important to you Now what? 15
  • 16.
  • 17.
    ● Do youever doubt the results your platforms are reporting? Why? ● How have you dealt with this before? ● Have you run a lift test? How did it improve your decision-making? ● For those who have not run a lift test, how could lift testing help your business? Discussion questions 17