5. More time spent daily than anywhere else
0
5
10
15
20
25
30
35
Facebook YouTube Yahoo! Google Twitter
Averagedailyminutes
desktop
Source: comScore Key Measures, US Desktop, December 2013
0
5
10
15
20
25
30
35
Facebook Google Instagram YouTube Twitter Yahoo!
Averagedailyminutes
mobile
Source: comScore Key Measures, US Mobile Metrix, December 2013
6. Clicks do not equal business results
Nielsen NetEffect meta analysis indicate
no correlation between CTR and ROI
No strong correlations emerge between CTR
and any of the Brand Effect metrics
Ad Recall Delta
Brand Awareness Delta
Message Consideration Delta
Purchase Consideration Delta
Brand Favorability Delta
-0.5 0.0 0.5 1.0-1.0
Click through rate
(correlation = -.07%)
0.10% 0.20% 0.30%
Breakeven
0.0%
1400%
900%
400%
-100%
ROI%
7. Reach
Did I reach the right
number of people and
the right type of
people?
Brand
resonance
Did I improve my
brand’s image and
change the attitudes
of customers?
Reaction
Did I cause a
customer
to act—to buy a
product, sign up for
a service or make a
referral?
Measuring against your business objectives
9. Studies show that Facebook can achieve
broad reach at far below the average cost
Source: Aggregate Knowledge, based on a representative sample of campaigns from Q4 2012. Reach represents percent of total cookies reached; bars include
overlap so will sum to more than 100%. Cost per unique user calculated by total cost of reach divided by total unique users reached by each publisher.
0%
10%
20%
30%
40%
50%
60%
Facebook Portal 1 Ad Network 2 Ad Network 3 Ad Network 4 Ad Network 5 Ad Network 6 Ad Network 7 Ad Network 8 Ad Network 9 Ad Network
10
Targeted
Network 1
Targeted
Network 2
Targeted
Network 3
Cost per unique user
Reach
Facebook
$0.001
$0.017
$0.009
$0.004
$0.008
$0.006
$0.012
$0.003
$0.005
$0.016
$0.007
$0.034
$0.009
$0.025
10. Facebook reach is highly accurate
Source: Nielsen OCR, October 2012
Online average
Facebook
Broad campaign accuracy Narrow campaign accuracy
94%77%
94%
Online average
Facebook
94%27%
91%
11. Facebook‘s large base of users
is simply not found elsewhere
Source: Aggregate Knowledge (across 10 billion impressions in Nov ‗12)
Facebook
is not ―stealing‖
conversions
77%
of FB reach cannot
be found elsewhere
Publisher 1
Facebook
Ad network 1
Publisher 5
Publisher 2
Publisher 4
Ad network 2
Publisher 6
Publisher 3
Ad network 3
Retargeter 2
Retargeter 1
Retargeter 2/
Ad network 5
Share of total reach (%)
Exclusivereach(%)LowHigh
Low High
12. Optimizing reach across platforms decreases CPM
Source: Nielsen MIO
$15.41
$11.98
$9.80
$8.29
0% 5% 10% 15%
Shift to Facebook
Total campaign: CPM
P18–49, Feb 2012
13. …without decreasing effective reach
Source: Nielsen MIO
43.6%
50.1% 52.3% 53.4%
0% 5% 10% 15%
Shift to Facebook
Total campaign: Effective reach
P18–49, Feb 2012
14. Measure cross-platform delivery with Nielsen XCR
TV only
TV and online
with Nielsen XCR
Combined audience
Online only
with Nielsen OCR
15. Facebook extends reach, complements TV
Note: Example, large QSR campaign, Audiences over 2 years of age
Male
audience
0%
20%
40%
60%
TV + Facebook
Facebook only
TV only
18-2015-172-14 40-44 45-49 50-54 55-64 65+ Total21-24 25-29 30-34 35-39
0%
20%
40%
60%
18-2015-172-14 40-44 45-49 50-54 55-64 65+ Total21-24 25-29 30-34 35-39
Female
audience
16. Total Facebook discussions vs. all other social media
5x more chatter on Facebook
15,706,665
12,093,555
10,275,888 10,233,971 9,985,681
8,328,015
6,505,076
4,598,372
3,747,028 3,509,402 3,234,482 3,142,618 2,939,642
2,249,807 2,213,801
7,078,632
2,632,384
132,742
2,428,133
415,063
1,750,395 1,427,817
1,768,891
140,028
872,361
170,533 2,217 176,776 259,690 140,737
Billboard Music AwardsCopa del Rey Soccer: Real Madrid vs. Atletico MadridMLB Baseball Yankees at OriolesNBA Playoff: Knicks at PacersStanley Cup Playoff: Rangers at BruinsNBA Playoff: Bulls at HeatNBA Playoff: Spurs at WarriorsAmerican IdolMLB Baseball: Tigers at RangersNBA Playoff: Grizzles at ThunderStanley Cup Playoff: Kings at SharksThe DoctorsStanley Cup Playoff: Blackhawks at Red WingsACM Presents: Tim McGraw's Superstar Summer NightNBA Playoff: Grizzlies bat Sp
FB chatter all other chatter
Source: Trendrr, 2013
5x
19. Greater brand resonance than
other online ad campaigns
Source: Nielsen Brand Effect norms, May 2012
Ad recall
Brand awareness
Facebook
Facebook
98%
online average
online average
31%
20. Research polls measure brand resonance
SPONSORED
Like This Page
Create Ad
RELATED
POSTSubway asked: It‘s Avocado Season!
Which sub with avocado are you going
to try first?
Turkey & Bacon Avoca
Chipotle Steak & Chee
BLT with Avocado
John Smith, Vikas Bhasin and Roslyn
Callahan Stewart like Subway.
21. Research polls reflect offline behavior
Source: Facebook internal polling & Box Office Mojo data, February 2011
Intent, Attendance & Weekend Gross
$0
$20
$40
$60
$80
$100
$120
$140
$160
$180
$200
0%
10%
20%
30%
11/5/10 11/12/10 11/19/10 11/26/10 12/3/10 12/10/10 12/17/10 12/24/10 12/31/10 1/7/11 1/14/11 1/21/11 1/28/11 2/4/11 2/11/11 2/18/11
Openingweekendgross(millions)
%intendingorreportingattendance
Intent: Are you planning to go the movies this weekend? R=.79
Attendance: Did you go to the movies this weekend? R=.89
Weekend gross: ($M)
22. Nielsen Brand Effect measures ad
effectiveness at driving brand metrics
1. Ad displayed to user
Users randomly assigned to exposed & control groups while ads are
served, yielding two groups that are perfectly matched on targeting
and site usage
2. Polls next day in ad context
• Summary results with demographic cuts delivered in 3 days (2 days post ad
run)
• High response rates, 10-100x of other methods, ensure representative sample
& not ―professional survey takers‖
Control group created dynamically
Representative results delivered quickly
24. OCM closes the in-store loop for your
Facebook marketing
―Facebook‘s new Custom
Audiences measurement could
make it clear that someone
didn‘t just stumble into a
physical store‘s big Sunday
sale, but instead saw an ad for it
that inspired their visit.‖
– TechCrunch, December 2013
25. Offline Conversion Measurement: Nuts and bolts
Test group
2
Holdout group
Target
audience
1
you’d like to
database
ed X
ed Y
ed in the last
FACEBOOKAD IMPRESSION
TESTGROUP
HOLDOUTGROUP
10%conversion =
5.Analyze ad impression data and transaction
data to calculate lift between two groups
FacebookDriven
2%conversion =
NOAD IMPRESSION
TESTGROUP
HOLDOUTGROUP
2.Split thisaudience into two groups
–test and holdout –and upload to FB
Client Driven
3.Target the test group with
media and exclude holdout group
Client Driven
4.Send FBencrypted transaction
data for audience to FBvia SFTP
Client Driven
• Conversion happened
• Conversion amount
• Conversion category
3
Test group
Holdout group
4Sales lift 5
CLIENT: Identify an
audience you‘d like to target
from your CRM
CLIENT: Split this audience
into two groups—test and
holdout—then upload to
Facebook
CLIENT: Target the test
group with media and
exclude the holdout group
CLIENT: Send encrypted
transaction data to
Facebook
FACEBOOK: Analyze ad
impression data and
transaction data to calculate
lift between two groups
1
2
3
4
5
26. Outcome measurement methodology
Propensity score matching
• Match exposed and unexposed users based on demographics, Facebook
engagement, fan status, etc.
• Compare ownership patterns across matched samples of exposed and
unexposed
Unexposed
to ads
Exposed to ads
Ownership determined
pre-campaign and post-campaign
Ownership determined
pre/post
27. Outcome measurement internal test
Analysis controls for gender, age, fan status, FB engagement level, etc.
Source: Internal Facebook data, 2013
Exposure frequencyWeeks from campaign end
Percentage lift in device adoption Device adoption by ad exposure frequency
wk 3 wk 4 wk 5 wk 6 wk 7
Control
Exposed
1 2 3 4 5 6 7 11 21 41 81
wk3
wk4
wk5
wk6
wk7
30. The opportunity: Optimize ROI with the right
message for right person at the right time
Grocery
product
Saturday
shoppers
Personal
care product
end-of-cycle
loyalists
Pet product
mid-cycle
Cleaning
product
recent
switchers
31. Moving towards a multi-touch attribution model
Provides a more realistic
assessment of ROI by capturing
a holistic picture of the customer
journey across channels
A multi-touch attribution model
attributes value to observed touch point
along the path to conversion
✖ Miscalculates ROI by assuming
the consumer doesn‘t see or click
any other ads along the way
A last-touch attribution model
attributes entire value of a conversion to
the last ad clicked or seen
32. Use MTA to identify optimal frequency
Source: Aggregate Knowledge (all campaigns in flight between Jan 1 and May 23rd 2012)
Frequency over 7 day period
1-2 3-5 6-
10
11-15 15-
20
21-
30
31-
40
50+
35
30
25
20
15
10
5
0
40
CPA
33. Multi-touch models reduce CPA
First test segment for this client:
Conquesting Fans of a
competitor
Overall multi-touch attribution is
promising (3-day view, 30-day
click windows)
Facebook View-Through Beta: Overall Results
$72.0
5
$110.88
$762.48
First- tough CPA Multi-touch CPA Last-touch CPA
34. People drive business results
33%increase in
actions, conversions and
sales
When clients moved from
allocating their advertising
spend using last-touch
attribution to using multi-
touch attribution, they saw a
Source: Aggregate Knowledge, Q4 2012
35. Exposure on Facebook
drives results
Source: Datalogix, June 2013
99
%
of people who saw a
Facebook ad and then
bought a product in the
store never clicked on
an ad at all
On average,
36. Reach your customers
more efficiently
70%higher offline
ROI than those
that didn‘t
Campaigns that maximize for
reach among their target
audiences achieved
Source: Datalogix
37. Identify
Find
Reach
Measure
Identify your target
customers in your CRM
Measure against your
business objectives
Reach them with relevant
Facebook ads
Find them on Facebook
using our privacy- and data-
protective match process
Create a test plan
to apply
learnings, repeat
and optimize
Incorporate learnings into future marketing
38. Summary
Measure against your Business Objectives
Optimise for Reach, Resonance and Reaction
Plan TV and Facebook media together
Use Multi Touch Attribution models to determine your media split
Optimise ROI with the right message, to the right people, at the right time
To understand where we are today, we’ve thought back to two of the major media platform developed in the last century, radio and TV. And when we look closely, we see a basic pattern in how measurement progresses. A new platform emerges, different players try to create ways to measure success and eventually a third party takes ownership of creating a success metric that can be applied and standardized across the industry. The GRP or gross rating point which allows television advertisers to know how much of their target audience they are reaching and at what frequency and became the standard buying unit for how advertisers buy television ad placements. This metric became immortalized once advertisers were able to plug this number into (media mix models (MMM) and show that the GRPs they were buying were driving offline sales in aggregate relative to other marketing methods.Online has struggled to create a standard metric like this across the industry. This is largely because online advertising is often bought in more precise and small budget allocations across fragmented sites that do not use similar metrics. Because of this, online has come to rely on bottom funnel, action-oriented metrics like the click-through rate.
In fact, we have seen this be validated time and time againPer Nielsen’s research, CTR does not actually correspond to the things marketers value (R, R, R). CTR does not correlate with standard brand metrics advertisers are trying to achieve when doing brand advertising.
And independent studies – like this one from Aggregate Knowledge – show that Facebook can achieve this massive reach at far below the average costIn this chart, across the campaigns that Aggregate Knowledge observed in Q4 2012, almost 50% of the cookies reached were on Facebook – and the cost per total unique user is a fraction of other publishersSo, in addition to incredible reach, we have the efficiency you’re looking for too
Reaching the right people at the right time with stories that are engaging and interesting to them sets the stage for brands to be able to achieve their brand objectives. According to Nielsen’s Brand Effect norms, on average, Facebook ads drive almost 2X (98%) better ad recall than non-Facebook campaigns measured in the same way.And, on average, Facebook ad campaigns drive 31% better brand awareness than non-Facebook campaigns measured in the same way.Note: The ads in this study are based on Nielsen’s Brand Effect database which includes a mix of ads with social context and those without.
Facebook is an incredibly powerful tool to collect real-time sentiment at scale. That’s because we are able to runs polls on the side of our Pages that feels like content so our response rates and samples sizes can be very large. These polls are all opt-in for facebook users.High response rates, users love them, light weight, quick response. Nevertheless, we still need to demonstrate that the data we collect are valid for the real world.
Movie sales was one of the first industries we were able to crack this for so we’ll use it as an example. As we talked about the accuracy of our polling product in predicting consumer sentiment and behavior, we worked with movie standard measurement company NRG to track intent to actual box office sales and found intent to be an excellent predictor of box office sales. The metric that is most important for the movie industry.
This allows allows us to conduct Brand Effect studies – our co-branded product with Nielsen. This methodology allows us to tie back direct impact of ad campaigns at driving the desired advertising results such as Ad recall, message awareness, & intent/considerationWe are able to run ads with a hold out identical sample and see the difference in perceptions across those who were exposed to the ad campaign versus those who were not.
So you probably see where we are going with this. Imagine that you have a new device launch and you have a fb campaign. I could look at adoption of the device at fixed time periods after the campaign ends, and then connect that to exposure to the ad. And we do some statistical analyses here (I can talk about this more but in addition to experimental randomly selected holdout we also utilize a statistical method called propensity scores matching which is the standard in the industry for making sure exposed are not different than control). And We add variables that we care about, like demographics, fan status, fb engagement, etc. and then we look at the lift in sales for exposed versus control. Now where is this now? We tested the methodology internally to make sure we have the right methodology and right ownership logic on a past campaign and the results both established the methodology and also showed 2digits percent differences in offline sales for exposed versus control. Now we want to push this out and partner with our closest partners, apply this to a relevant campaign, get the insights and feedback from our partners about the methodology, and then hopefully be able to push this forward. Where we want to be in a year is move beyond the beta stage and apply this as a standard, with the goal of helping you understand which campaigns were better – so the goal would be to get this done repeatedly and help you understand what makes your campaigns on fb better, and how do they compare in their effectiveness to tv and other media platforms. For Telco, we just came up with a unique methodology to measure ROISo, imagine that you have a campaign, and just like what we do for brandeffect, we will create a control and treated group. Specifically, we will randomly pick a certain percentage of your target audience that logs into facebook to not be exposed to your ad. So we create a hold out or control group that looks exactly like the exposed group and is chosen randomly. Based on this, we can then compare churn rates out of O2 and into O2 in the control versus the hold out group and examine whether the campaign is effective.
So you probably see where we are going with this. Imagine that you have a new device launch and you have a fb campaign. I could look at adoption of the device at fixed time periods after the campaign ends, and then connect that to exposure to the ad. And we do some statistical analyses here (I can talk about this more but in addition to experimental randomly selected holdout we also utilize a statistical method called propensity scores matching which is the standard in the industry for making sure exposed are not different than control). And We add variables that we care about, like demographics, fan status, fb engagement, etc. and then we look at the lift in sales for exposed versus control. Now where is this now? We tested the methodology internally to make sure we have the right methodology and right ownership logic on a past campaign and the results both established the methodology and also showed 2digits percent differences in offline sales for exposed versus control. Now we want to push this out and partner with our closest partners, apply this to a relevant campaign, get the insights and feedback from our partners about the methodology, and then hopefully be able to push this forward. Where we want to be in a year is move beyond the beta stage and apply this as a standard, with the goal of helping you understand which campaigns were better – so the goal would be to get this done repeatedly and help you understand what makes your campaigns on fb better, and how do they compare in their effectiveness to tv and other media platforms. For Telco, we just came up with a unique methodology to measure ROISo, imagine that you have a campaign, and just like what we do for brandeffect, we will create a control and treated group. Specifically, we will randomly pick a certain percentage of your target audience that logs into facebook to not be exposed to your ad. So we create a hold out or control group that looks exactly like the exposed group and is chosen randomly. Based on this, we can then compare churn rates out of O2 and into O2 in the control versus the hold out group and examine whether the campaign is effective.
Advertisers that have used Offline Conversion Measurement to measure the impact of their marketing in News Feed realized a median ROAS of 8X
Your attribution model mattersWhat many people use today is a “last-touch attribution model”But last-touch attribution is like watching a soccer game and only seeing the shots on goal – the last kickInstead, we recommend a multi-touch attribution model approachWith a multi-touch model, you can see all of the players on the field for the first time – understand which players are making the assist and which ad buys are contributing to conversionAnd, isn’t that something you’d want to know?We have lined up several multi-touch attribution firms to help you get this right:Aggregate Knowledge, Convertro, Adometry, VisualIQ, ClearSaleing, KenshooFacebook has no formal relationship with these partners – they are on your side to help you evaluate your marketing activities on Facebook and across all of your digital ad buysUsing view and click tags, these attribution partners can generate a holistic view of all of the touch points to your customers on the path to conversion and report on and which ones are driving the actual conversion event
The chart:Vertical axis is CPA. Lower CPA is better.Horizontal axis is frequency bucket. Somewhere in the middle is optimal, but it will vary by client.The key message: Aggregate Knowledge (using view tags on your paid media) enables you to build a holistic view of frequency. Even when you optimize frequency on each publisher, across publishers you can drive up frequency to 50+ which increases your CPA. The reason is that many ad networks, publishers, and retargeters serve your ads against the same pool of cookies that you’ve seen elsewhere. What you should care about as an advertiser is the unique reach across publishers.
Multi-touch here is just a simple even distribution. If there were seven ads seen along the path to conversion, each gets 1/7th credit for the conversion. This is not the most rigorous approach, but it simplifies the conversation.
Why is it important to measure people instead of clicks? Because people buy products and clicks don’t correlate with success.Facebook’s ability to connect advertising exposure to actual business outcomes, online and offline, is truly unique. This capability provides more than just ample evidence that Facebook works, it allows us to provide you with deep insights on your customers and how marketing on Facebook drives results. Our ability to measure real people, not cookies or clicks, allows you to measure and optimize your investments throughout the purchase funnel -- whether your goal is to reach new audiences, launch a product, or increase online sales or foot traffic (or any combination of these). Most importantly, Facebook often drives higher return on ad spend than other media.
Why is it important to measure people instead of clicks? Because people buy products and clicks don’t correlate with success.Facebook’s ability to connect advertising exposure to actual business outcomes, online and offline, is truly unique. This capability provides more than just ample evidence that Facebook works, it allows us to provide you with deep insights on your customers and how marketing on Facebook drives results. Our ability to measure real people, not cookies or clicks, allows you to measure and optimize your investments throughout the purchase funnel -- whether your goal is to reach new audiences, launch a product, or increase online sales or foot traffic (or any combination of these). Most importantly, Facebook often drives higher return on ad spend than other media.
