MAU Vegas 2016 — How to Build a UA Infrastructure to Maximize Your Time and E...
Mau 2015 scaling facebook at care.com
1. Wei
Kuan
Lum,
Sr.
Director,
Acquisi6on
@
Care.com
Scaling
Facebook
By
Hand
2. #MAU2015
The World's largest online destination for care. We connect families with great
caregivers and caring companies to help you be there for the ones you love.
Care.com
4. #MAU2015
Roadmap…
• Campaign Development
• Definitions 101
• A View of the FB Acquisition Curve
• Campaign Types & their Value on the Curve
• Look-a-Likes
• Interest Group Targeting
• Video Ads
• Creative Hacking & Management
• Training the FB Algo
• Speed Bumps
• Wrapping it Up…
5. #MAU2015
NavigaDng
the
OpDons
TargetingBidding
Options
Creative
Structure
oCPMCPI/CPA
Click/Views
IG
WCALALs
How many?
ButtonsStyle
Campaign
AdsAd Set
6. #MAU2015
TargeDng
101
–
DefiniDons
Look-a-likes Interest Group Website
Custom
Audiences
Email List Uploads
Nest the Audience
- 10-9%
- 9-8%
- Etc…
Keywords used to
ID existing & new
audience traits
Retargeting Pixel
Video Ads
Drive Views &
Conversions
Keep Audiences >1MM, “Perfect Audience” size varies by Vertical
7. #MAU2015
Facebook
–
Campaign
Acq
Curves
Interest Group Targeting
Lookalikes
WCA
Video Retargeting
CPA/CPI
Conversions
8. #MAU2015
LALs…
Exclude,
Exclude!
Look-a-likes
Email List
• Slice & Dice
Exclude, Exclude!
Final Audience Size is
Key
• Keep It Broad (e.g. Age
Range, etc.)
• Unless Audience is too
Large (e.g. 20MM)
• Age Range
• Actives vs Inactives
• Engaged
• Returning
• Churned
• Vertical/Product
• Keep Lists Fresh –
Timeframe!
9. #MAU2015
Interest
Group
–Audience
Developm’t
What Do I Like to
Read?
How Do I Relax?
What Are My
Hobbies?
Who Do I
follow?
Interest Group
Psychological
• Brands
• Politics
Social
• Likes
• Spend Time
• Events
Emotional
• Book types
Practical
• Products Used
10. #MAU2015
LALs
+
IGs=
Incremental
Volume…
Exclude,
exclude!
Look-a-likes
Email List
• Slice & Dice
Exclude, Exclude!
Final Audience Size is
Key
• Keep It Broad (e.g. Age
Range, etc.)
• Unless Audience is too
Large (e.g. 20MM)
Psychological
• Brands
• Politics
Social
• Likes
• Spend Time
• Events
Emotional
• Book types
Practical
• Products Used
Interest Group
Watch & Prune: Targets
Can Compete
11. #MAU2015
IG
TargeDng
–
SOPs
Interest Group
Test: Place the Targeted Individual
into “Interests” section
12. #MAU2015
Video
Ads...Increment
the
Curve
Video Ad
Capture New Audience
• Views = Qualification
Retarget Video Viewer
• Keep It Broad (e.g.
Age Range, etc.)
• Unless Audience is too
Large (e.g. 20MM)
• -10%-+15% CPA
13. #MAU2015
CreaDve
–
Hacking
for
Scale
How Well Can Your Target Do?
Creative is King… So test, learn, Rinse & Repeat
• Creative determines the ability for a target segment to scale.
• Each creative has it’s own “reachable audience” so many cases, campaigns
need multiple creatives
• One Size Does Not Fit All: Creative fatigue varies by vertical
• Make it Relatable! (To IG Target, To Audience)
Optimize budget for creative
Reach = X
Reach = X+Y
Reach = X + .5Y
14. #MAU2015
Managing
CreaDve
TesDng
Color Filter Messaging Style Image Type Style of Image
Filtered • Syntax (Punctuation) # of items Person in Action
Native • Sell Message Types # of people Person + Product
Bright • Short Sentences Wilderness, Store, Park
Dark • Title Case Logo vs Not
Multi-Product Ads… A Few Thoughts
• Awesome CTR Play
• MPAs set the new bar for CTR threshold and Auction Eligibility
• It’s a Balance!
• Skew towards CTR? CPA/CPI is too high
• Skew Targets CVR? Deliverability Suffers
15. #MAU2015
Training
the
FB
Algo
Campaign Structure
• Naming Convention –Determine Performance W/o the Double Click
• Platorm_DEMO_IG_Targets_Other Target_Date
Choosing Bidding Types
• Going for scale? oCPM
• Is Scale is penultimate to Budget? CPI
• Not both
Pixel Placement? Provide Sufficient Signal
• Less Optimized Funnel?
• Choose a Point Higher in Funnel, Measure on Backend
• Short, Highly Optimized Funnel?
• Choose an Optimization Point Downstream
16. #MAU2015
Determining
Budget
Setting Initial Budget Impacts Performance Dramatically…
Pre-Work to Campaign Creation…
• Think About Your Businesses: What is my allowable CPA?
• How do I give FB budget to run & provide the algo sufficient signals to remain valid?
• Don’t Go to Big or Too Small with Budget
• Enable FB to Receive the Right Signals at the Right Cadence!
• Monitor.. Monitor…Monitor!
17. #MAU2015
Possible
Speed
Bumps
Algo is Sensitive & Still in Development:
• Finicky campaign history results in slow and very manual rebound:
• Be Cautious: Campaigns and ad “death” occurs within hours of a campaign or ad
being turned off, without ability to regain prior high performance.
• FB is Not Google: One cannot turn off a high-performing ad and expect it’s prior
efficacy or scale to return
• With delayed UI, misfires with top campaigns = very manual ramp back
• UI sometimes locks an advertiser out from building campaigns
Negative feedback loops Exist!
• Ad/Campaign CVRà influence account conversion rate à reduces conversion
rate across other campaigns.
• Optimizing towards different conversion points in same portfolio can cause competition
and set a new conversion rate
Fine Balance between CTR and CVR
18. #MAU2015
Possible
Speed
Bumps
Fine Balance between CTR and CVR
Key Challenges/Learnings
“Instablity”
• Interest Group targeting can fatigue quickly, and without notice
• “De-dupes”: IG can compete, even if they once did not (What’s hot today changes).
Cannibalization occurs, resulting in volume tanking
• Make sure creatives are great but not too awesome. (cannibalize other ads)
UI can be very delayed or not work,
• Less than optimal decision-making
19. #MAU2015
In
Summary…
• Be Willing to Get Crazy… Be Obsessive
• Be Creative, Leave No Stone Unturned
• Be Patient
• Have Grit
• Questions/Comments/Keep the Convo alive?
• LinkedIN Profilel:
https://www.linkedin.com/in/weikuanmarketinganalytics
• Email: weikuan.lum@stanfordalumni.org or wlum@care.com