Facebook’s IDDM

Playbook

MARCH 2020
IDDM can increase the campaign results by adding relevancy

DEFINITION Individualised Data-Driven Marketing (IDDM)

Using digital data signals on individuals & context to target the right people with
relevant (personalised) messaging

IDDM helps drive Reach… … and Resonance … … which improves Reaction

Right people at the right Right message, personalized More sales & improved ROI
time and place using relevant available data from the same Marketing
triggers budget

2
However, a wrong execution will cancel all potential benefits

iDDM main pitfalls: Solutions:

• Audiences that are not Meaningful,

Addressable and Predictive (M.A.P.), more
• Define the most appropriate
specifically: Personalization strategy based on
your business needs and
• Audience segments built without
insights or data signals won’t generate capabilities (building audience
stronger reaction segments is not the unique way of
• Over-narrowing audiences, drastically
doing iDDM)
reduces the Reach and therefore will
increase CPMs (smaller audiences are • Machine Learning delivery instead
harder to find by our Delivery system) of manual targeting
• Overlapping audiences: unknowingly
decrease the Reach and increase the • Dynamic ads (w/ 3P Partners)
Frequency and CPMs instead of producing 100’s of assets
• Creatives not personalized enough: Creative
should be consistent but unique to provide
incremental relevancy

3
Personalization at scale: Same costs and reach, stronger reaction

FROM… …TO Each person get served

a personalized ad based
on their preference in
order to maximize your
campaign impact
Creative A

Creative B
Broad approach,
generic creatives
Creative C Creative D

10M Reach 10M Reach

M/F, 18-55 M/F, 18-55

4
IDDM Checklist

1) Make sure your campaign plan is fully EFE compliant See slides 7-9

2) Choose the right Personalization strategy based on

See slides 10-22
your business needs and capabilities

3) Set-up the right measurement solution See slides 23-27

1

‘EFE’ still applies to

IDDM campaigns
IDDM Media Best Practices (EFE 2.0)
OPTIMAL PLANNING GUIDELINES

METRIC DESCRIPTION NON-OPTIMAL MEDIUM RESULT OPTIMAL RESULT

% of campaigns that are achieving a +1 weekly Average frequency of less Average frequency of more
Strategic frequency
frequency than one per week - than one per week

Campaign duration % of campaigns longer than 4 weeks Less than 4 weeks - 4+ weeks

Optimal results
proven by
% of campaigns that are reaching at least 40% of their
% of audience reached Less than 30% reach 30%-40% reach 40% - 60% reach Heineken’s Meta
target audience analysis
OPTIMAL BUYING GUIDELINES

What are the campaign objectives? Non-optimal objectives

Objective BAO / Video views Reach
(% of OpCo spend) (PPE, Traffic)

No placement optimization PO with FB & IG Feed and

Placement optimization % of campaigns enabling Placement Optimization PO with FB & IG Feed (2)
(PO) Stories (4)

Buying Type Are the campaigns purchased in Auction or R&F? (% of - - Reach and Frequency
OpCo spend)

For visibility
purpose only. The
Campaign budget % of campaigns enabling Campaign Budget
No - Yes solution can be
optimization Optimization
adapted to the
campaign needs
IDDM Creative Best Practices (EFE 2.0)

MANUAL
Business Objective Instant Attention Frame for Mobile
SCORING OF
Is the creative tied to a Does your image/video have Is the asset framed for mobile
singular business objective and strong stopping power or grab feed/Stories? 1:1 to 4:5 for TOP 10 HIGHEST
tackling the intended benefit? attention in the first 1-3 secs? Feed, 9:16 for Stories? SPENDING ADS
PER COUNTRY

Think sound off…and Brand Link

sound on Most frequent
Does it incorporate Brand non-compliant
Is the core message clear identity from second 0? criteria Q2 ‘19 Q3 ’19
without sound? Does it delight
with sound? Business Objective 0 1

Instant attention 5 9

Retain interest 8 11
Retain Interest
Highlight what Matters
Brand link 8 12
Does your video retain
Are you communicating attention via storytelling or
your main message in the Designed for Sound off 7 2
other techniques?
first 1-3 sec? Frame for Mobile 7 13

Highlight what matters 14 7

8
2

IDDM Personalization
strategies
IDDM Personalization strategies
Choose the right Personalization strategy based on capabilities and business needs

High Signals Capability

Contextual Individual
Relevance in the moment Relevance to Me
e.g. Product, Context, Weather, Sports, Fixed event e.g. Pixel Data, CRM, Collaborative ads
dates/time of day
Lower Personal Higher Personal
Relevance Relevance
– –
Higher scale Lower Scale

Demographic Interests
Relevance to Demographic segments Relevance to Personas
Mutually exclusive audience segments with Overlapping audiences with Automated Targeting
Manual Targeting

Low Signals Capability

Which approach can you apply?

No 5
Can your brand leverage the
Do you have a customer Can you leverage contextual Generic
segmentation that you want to signals (Location, time, day,
Facebook Pixel or CRM data?
No apply? No event…) in your creatives?
campaign
Yes
Yes Yes

Does the segmentation only

use demo variables
(e.g. gender, life stage)? No

Yes
1 2 3 4
Individual Demographic Interest Contextual
Personalization Personalization Personalization Personalization

Targeting: Targeting: Targeting: Targeting:

• FB Pixel retargeting and • Manual targeting of • Automated targeting, • Automated targeting,
look-alike audiences mutually exclusive demo with all creatives to a with all creatives to a
based on CRM audiences broad audience broad audience

Creative: Creative: Creative: Creative:

• Facebook Dynamic Ads • A few distinct creatives • A few distinct creatives • Several distinct creatives
optimized for conversions personalized each for a personalized each for a personalized based on
specific segment specific segment contextual signals
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Individual Personalization strategy

Creatives delivered dynamically based on Data signals to the right user at the right time

Pro’s: Most accurate targeting, based on individual

Use the Individual Personalization strategy to increase ROI of behaviour
When to use it? lower-funnel campaigns (purchase intent, online sales) by
leveraging strong data signals Con’s: 1P/3P customer data will be difficult to
gather at scale

Data requirements: Leverage individual data such as Pixel data, CRM, Video views or 3P data to retarget or build Lookalike audiences
Requirements
Creative requirements: Personalized creatives built for the specific objective (DR or Brand?) and targeted audience behaviour

EFE guidelines (only applicable for Brand campaigns): 1 Campaign (Auction)

Min 4 weeks duration
Campaign set-up Min 1pw frequency
Reach between 30-60% of target audience Retargeting or Lookalike based on Pixel, CRM or 3P data
Creatives “7/7”
Dynamic content delivery
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Individual strategy:
Campaign examples
DEUTSCHE BAHN HEINEKEN USA
Deutsche Bahn – Save Yourself a Flight Link Ad (DTC) Collaborative Ads
72% of Germans go on vacation abroad. How do you get them to travel to Germany with
Deutsche Bahn? Using an algorithm that compares side-by-side pictures of popular holiday
destinations with their German equivalent, and pulling dynamic flight prices based on the
users’ closest airport.

Location – Ads and prices adapted based on the user location and it’s closest airport

Behaviour – The campaign targeted users who expressed a Travel interest.

Dynamic Ads – The video ads were built based on the available destinations and their German counterpart, with a
dynamic flight price displayed. Images were regularly updated to fight creative fatigues

HUSA – Collaborative Ads

Heineken USA leveraged e-retailers platforms (Minibar and Drizly) to run e-commerce Ads. By
sharing its Pixel data (website traffic, Heineken purchases) and their online Heineken product
catalogue, the e-retailer allows the advertiser to send qualified traffic to their e-commerce site,
optimized for purchases.

Pixel data – The advertiser can run e-commerce ads without a pixel or online catalogue, by leveraging the e-retailer’s
pixel data

Dynamic Ads – The product shown will match the user preferences, leveraging its browsing and FB/IG behaviours.
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Demographic Personalization strategy

Personalized creatives targeting mutually exclusive audience segments based on Living Location, Age,
Gender or Life-stage.

Pro’s: leverage clearly addressable data signals. Add

Use the Demographic Personalization strategy if you identified relevancy without sacrificing the scale
When to use it? or want to prove that different demographics will react better
to a creative personalised for them Con’s: Increased cost of production and costs of
delivery due to multiple audiences

Data requirements: None. Leverage FB data

Requirements
Creative requirements: The video must be shot with the different customers (gender/age) or representing the different locations

EFE guidelines : 1 Campaign (R&F or Auction)

Min 4 weeks duration
Campaign set-up Min 1pw frequency
Reach between 30-60% of target audience
Audience 1 Audience 2 Audience 3
Creatives “7/7”
Creative 1 Creative 2 Creative 3
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Demo strategy:
Campaign example Mutually exclusive, collectively exhaustive

EDUCATION EDUCATION
Benefit Cosmetics leveraged the insight that women of different 18–34 Carousel 35–60 Carousel
age have different needs when it comes to brow cosmetics.

They built 2 Video Carousel Ads highlighting distinct benefits and

delivered them to 2 mutually exclusive segments.

Ad Recall
+14 1.5x CPG Norm 9

Message Association
+9 9x CPG Norm 1

Purchase Intent
+10 10x CPG Norm 1
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Interest Personalization strategy

Personalised creatives targeting interest-based audiences (with overlap between them).

Pro’s: Apply a flexible and granular segmentation, using

Use the Interest Personalization strategy if you identified or Automated (Machine Learning) delivery to work around
want to prove that Interest-based audiences (personas) have the audience overlap
When to use it? different needs or will react better to a creative personalised
for their interests. Con’s: Creative assets should be significantly
different to drive incremental impact

Data requirements: None. Leverage FB data

Requirements
Creative requirements: Personalized creative assets should be drastically different to drive incremental impact

EFE guidelines : 1 Campaign (R&F or Auction)

Min 4 weeks duration
Campaign set-up Min 1pw frequency
Let FB’s Machine
Reach between 30-60% of target audience Broad audience combining all segments Learning deliver
the right
Creatives “7/7” creatives to the
right user
Creative 1 Creative 2 Creative 3
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

FB ML combines data and signals from our

What is Machine platform, with insights you share, in order
to make predictions for who the right
Learning? people are for a given message.

As people take different actions on and off

Facebook, it creates intent signals that help
us deliver a more tailored ad experience.

Personalization is built in
everything we do at Facebook.
We personalize over 2.5B Feeds
every day
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Benefits of Machine Learning Delivery

Eliminates need for Eliminates need for Eliminates concern Maximizes creative
mutually exclusive mutually exclusive over smaller effectiveness
audiences audiences audience segments

Manual segmentation would Manual segmentation would Manual segmentation resulted ML maximizes strengths of
force people into 1 segment, result in wasted impressions in concerns around scale and assets both at the audience
even if they have higher due to duplicated reach and cost of smaller, more niche level and at the creative set
propensity for another frequency across segments audiences, creating CPM level.
segment. This allows ML to do inefficiencies and
that delivery for you. ML maps the most effective measurement/optimization
creative to most effective black holes.
audience using our FB
Algorithm.
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Interest strategy:
Campaign examples
HEINEKEN KOREA GUINESS MALAYSIA
Several Heineken tests have proven that letting the Machine Music Lovers Ad Sports Ad
Learning do the creative allocation (= all creatives under 1 Ad set)
drives stronger ROI than doing it Manually (= 1 Ad set per
audience).

South Korea Results

4 Audience segments based on Interests:
• Travelers Efficiency of Ad Recall for
• Foodies
+63% the Automated Targeting
• Music Lovers Cell
• Home Entertainers

Malaysia
+63% Efficiency of Ad Recall for
3 Audience segments based on Interests: +5% the Automated Targeting
• Work
Cell
• Sport
• Food
Impressions for the
+7% Automated Targeting Cell
INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Contextual Personalization strategy

Creatives personalized on contextual elements targeting a broad audience.

Personalized creatives for lunch and Personalized creatives for weekdays, Personalized creatives based on the Sport or Use 3P Partners to develop Dynamic Ads
Time of day Day of week Location
evening (generic creatives rest of day) weekends, or game days user location Weather based on external live data feeds

Pro’s: Personalization with no added delivery cost. Contextual

Use the Contextual Personalization strategy if you identified or triggers are relevant to the beer category (temperature, time/day,
When to use it? want to prove that adding a contextual element to your sport)
creative will drive higher impact. Con’s: Need to work with a 3P partner to add live data feed
(weather, sports)

Data requirements: Use FB data + 3P Partner if live data feeds are needed (weather, sports results)
Requirements
Creative requirements: Personalized creatives for each situation + 3P dynamic creative if needed (weather, sport results)

EFE guidelines : Time of day / Day of week Location Sports or Weather data
Min 4 weeks duration
Min 1pw frequency 1 Campaign (R&F) 1 Campaign (R&F or Auction) 1 Campaign (Auction)
Campaign set-up Reach between 30-60% of
target audience 1 Ad set (Broad audience with scheduled
Location 1 Location 2 Location 3 1 Ad set (Broad audience)
Creatives “7/7” delivery)

Dynamic content delivery Creative 1 Creative 2 Creative 3 Dynamic Creative

INDIVIDUAL DEMOGRAPHIC INTEREST CONTEXTUAL

Contextual strategy:
Campaign examples
McDeliverain Flywheel
McDonalds Paris – McDeliverain Feed Video Ad Feed Video Ad
Leveraging weather-based targeting, dynamic movement, and one-click conversion,
McDonald's meaningfully connected with French Millennials at their rainy moment of
hunger resulting in 20,969 app downloads, +6.5pt in order intent and a CPI of 0.48EUR.

Weather – Ads appeared in feed/stories the moment rain began

Location – Hyper-targeted to postal codes in France so when it rained on windows, it rained in phones

Time – Ads focused in on lunch & dinner day-parts when people were most likely to be hungry/searching for
inspiration on social

Heineken UK – Flywheel
Project Flywheel has proved the value of digital marketing in pubs, bars and restaurants.
Activating in over 6,000 outlets across 10 different brands with Facebook delivering the highest
ROAS and becoming the sole platform for delivery. The campaign delivered a £2.2M incremental
sales over 18 months with a £10.04 ROAS for the best performing cells.

Location – Hyper-targeted user exact location

Time – Ads focused in on lunch & dinner day-parts when people were most likely to be thirsty/searching for
inspiration on social

Dynamic Ads – 1000’s of creative assets (for each pub / brand combination) built and delivered using Spirable
3

IDDM Measurement
solutions
IDDM execution tests

Business question Testing approach

Creative build for mobile: Which personalization strategy (Demo, Interests, Contextual, Individual) drives
RESONANCE
What is the most effective personalization strategy? the highest value?

Targeting: Manual v automated targeting

REACH & Measuring the impact of manual allocation of creatives to audiences v
What is the most effective approach to link a personalized creative to an
RESONANCE automated delivery of personalized creatives to a broad audience
audience?

Solution Minimum spend Creative Duration Frequency Reach

• Multi-cell See market minimums All IDDM creatives must be meaningfully different from each other 4 weeks+ 1pw+ 30-60% of total
Facebook Brand target audience
Lift All IDDM creatives must follow eFe 2.0 creative brilliant basics principles:
• MMM / CMM • Instant attention
• Heineken sales-lift • Framed for mobile
study • Thumb stopping power
• 3rd party sales-lift • Brand in the 1st second
study • Message in the 1st second
• Optimized for sound off
Creative: What is the most effective Personalization strategy?

Demo strategy (Who?) Interest strategy (Who, What?)

Demographic or Non-overlapping audiences with Personalised creatives targeting Interest-based
specific creative tailored to their needs audiences (with overlap between them)

Recommended test structure:

M.A.P M.A.P M.A.P • Test one Personalization strategy
M.A.P. Broad audience (Cell 1) against another (Cell 2). If an
Audience 1 Audience 2 Audience 3 (Let FB’s Machine Learning deliver the right creatives to the right user )
alterative personalization strategy is
not relevant, add a generic cell as a
Creative 1 Creative 2 Creative 3 Creative 1 Creative 2 Creative 3 second cell for your test
• If you are able to run a 3-cell test,
add a generic cell as well

Contextual strategy (Where, When, Why?) Individual strategy (5Ws)

Dynamic Creatives based on either location, time Creatives delivered dynamically based on Data
of day, day of the week or 3P data sources signals to the right user at the right time

M.A.P. Broad audience M.A.P. 1P or 3P data audience

Dynamic content delivery Dynamic content delivery

24
Manual v automated targeting: When does automated targeting drive
better outcomes v manual targeting?

Cell 1: Manual targeting Cell 2: Automated targeting

Let the system optimize the delivery of the creatives to a broad
Manually assign a creative to an audience (Campaign Budget
audience (Campaign Budget Optimization is not applicable)
Optimization is off)

M.A.P M.A.P M.A.P Combined audience from Cell 1 or Broad audience

Audience 1 Audience 2 Audience 3
Same creatives as Cell 1
Creative 1 Creative 2 Creative 3
Creative 2 Creative 3 Generic
Creative 1 (optional)
• This cell could be Demo or Interest type of personalization strategy
• The hypothesis that consolidating an audience will deliver better
business outcomes is based on two premises:
• Efficiency: By consolidating an audience we will achieve a more
cost-efficient delivery
• Effectiveness: Several creatives can resonate with the same
person therefore by not manually allocating a creative to an
audience we will leverage the machine learning algorithm to
deliver the most relevant creative to the right person
• Although we do not recommend doing manual targeting when
implementing Interest strategy, this test will help measure the efficiency
driven by automated targeting and audience consolidation
25
Thank you!