1. Big Data for Social, Search and
Online Display
Kevin Lee, CEO, Didit.com
Yann, Tanini, VP, E-Commerce,
Redcats Group
2. Founded Didit in 1996, Are we an
Agency or a Managed Campaign
Technology, or a Tech-Company?
Written three books, over 400
columns/articles spoken at 450+
conferences/events.
2007
2007, 2008, 2009
Didit’s focus / differentiation: After 17years, it depends who you ask.
Some would say service others
“one-stop-shopping” & yet others
love the technology.
In the end its value delivered.
@Kevin_Lee_QED
3. The Many Faces of Big Data
“Big Data” or the application of analysis on large
data sets can be powerful at various stages of
marketing and advertising as well as within the
enterprise.
•
Paid Media
•
Earned/Shared
•
Site-Side
•
CRM/SFA
@Kevin_Lee_QED
4. Marketing & Media, a game of odds
You want to do things that move the odds in your
favor. Big Data helps:
•
Increased conversion
•
Larger purchases (immediate)
•
More profitable immediate purchases
•
Higher LTV (repeat conversions)
•
Likely to be an influencer
•
Likely to convert offline
@Kevin_Lee_QED
5. Attribution Modeling?
Even better: Marketing Mix Modeling
User served display
Ad
User Searches
User served display
Ad
User Visits
Website
User Converts
USER
Marginal elasticity of every media option is key!!
e·las·tic·i·ty - ilàstíssətee
In economics, elasticity is the ratio of the percent change in one variable to the percent change in another variable. It is a tool for
measuring the responsiveness of a function to changes in parameters in a unit-less way.
Elasticity is one of the most important concepts in economic theory. In addition when understanding how to allocate
media dollars, nothing beats knowing the elasticity of the market in every direction.
Touch Point Analysis is only the tip of the iceberg.
@Kevin_Lee_QED
6. “Big Data” improves results
What data do you have access to that:
• Predicts conversion rate
• Predicts average order
• Predicts lead quality
• Suggests personalization
• Predicts Lifetime Value
• Predicts influencers (social media or offline)
• What other KPIs would you like to predict?
@Kevin_Lee_QED
7. Data you might want to collect
What data are you now throwing away that could
be stored in a cookie or BI platform?:
• Search Keyword / Engine
• Geography / IP address
• TOD / DOW
• Visitation Count (annual)
• Operating System / Device
• and dozens more
@Kevin_Lee_QED
8. Where does your data sit?
Your own data + third-party data. Killer combo.
• CRM systems
• Business Intelligence (BI)
• Sales Teams
• Customer Service Teams
• Common Sense (might not work, but it may
work great in concert with existing data)
@Kevin_Lee_QED
9. Search Engines & Social Media
Search 3.0 is the intersection of search and social.
Additional relevance signals extend beyond your site
/ page to include people (the author of content or a
social media persona for example).
These additional signals are a big driver of “semantic
search” and search intent.
@Kevin_Lee_QED
10. Big Data and Content / Messaging
There is data out there about what content
will likely work for your business. Same for
messaging.
Boring doesn’t go viral.
Boring doesn’t get links.
Boring landing pages don’t convert.
@Kevin_Lee_QED
11. Big Data for SEO: Fun Links = branding
The better SEO strategies extend the content
strategy beyond the immediate prospects.
Big Data suggests the hot topics.
Fun content gets link juice, even if the visitors
aren’t prospects at that specific point in time.
My wife’s site ranked #1 for “dating mistakes” for 5 years (till site hack).
@Kevin_Lee_QED
12. SEO - Social Media: Fun Links = branding
Smart SEO recognizes the value of site visitors
at every stage of the buying cycle.
Smart Social Media strategies share that view.
Repeat exposure is key.
How can you use Big Data to measure if
branding exposures are working?
@Kevin_Lee_QED
13. Good Social Media delivers links = SEO
If the social media ecosystem is talking about
you, your product and your story, some of
that content will be written about in venues
that deliver links (and therefore link juice) to
your site.
Authentic links are the best for SEO. Social
media helps generate authentic links.
@Kevin_Lee_QED
14. Social Signals as SEO Validation
Several research studies seem to indicate that
social media signals are also being used as a
validation method to confirm that some pages
and sites deserve to rank.
Tie breaker between sites than all have great
content and could rank top may be social
media signals.
@Kevin_Lee_QED
15. Big Data in PPC Geo-Segments
Look Familiar?
The last year made it clear. Not all geographies share
the same demographic.
@Kevin_Lee_QED
16. Attributes odds change by geo
Demographics, Political donation levels by party
Psychographics
Personas
Real Estate values
InfoScan, Magazine readership, Internet access
speeds
Apple / Mac / iPad penetration
Use your imagination
@Kevin_Lee_QED
17. Geographies
Beverly Hills vs. West Hollywood?
• Upper East Side of Manhattan vs. East Harlem
• New Haven vs local community(zip code)
• Think about where you live, diversity by
geography
• Business density even works to tune audience
for some B-to-B
@Kevin_Lee_QED
18. The Power of Audience
Keywords tell you very little about audience
Geography predicts demographic and psychographic
audience segments:
• Conversion rate
• Income
• Wealth
• Weight / BMI
• Age
• and dozens more
@Kevin_Lee_QED
20. What's your Avg. Position?
For non-branded keywords - your avg PPC position?
@Kevin_Lee_QED
21. Conversion & Profit Lift Deliver Leverage
Conversion (and more profitable conversions)
deliver leverage.
Conversion/Profit =Leverage
Benefit of Leverage manifests itself as an increase
in reserve bid price (may not be fully needed).
Better Audience Delivers Bid
Leverage! Big Data delivers audience
@Kevin_Lee_QED
22. Geotargeting:
Conversion Rate Differences
- Average conversion
rate for US: 1.5%
- San Diego
conversion rate:
2.1%
You can bid over 25% more for
San Diego clicks.
What drives this change in conversion %?? AUDIENCE!!
@Kevin_Lee_QED
23. Geotargeting: higher spenders
Miami DMA shopping cart: $178
Average shopping cart: $122
→ you can bid 45%+ more for
Miami metro clicks.
What drives this change in conversion %?? AUDIENCE!!
@Kevin_Lee_QED
24. Geotargeting: higher LTV
Chicago DMA customers order more
often & longer tenure
Average orders over 3-years: 10
Chicago DMA orders over 3-years: 15
You can bid 50% more for new Chicago
customers.
What drives this change in conversion %?? AUDIENCE!!
@Kevin_Lee_QED
25. Geo is better with optimal ads
Ad units make it easier to talk to your largest profitable
segments differently, and optimally:
1. Separate ads by geography
2. Promotional language, offers
3. Different landing pages
4. Don’t rely on Ad Extensions
But what about enhanced campaigns?
@Kevin_Lee_QED
26. Geo Bid Boost using Big Data (govt+)
AdWords
$
$
$
$
$
Demographic MAP Data
$
$
$
$
$
Interfaces with AdWords account
• Determine Geo Bid Multipler based on demos
• Selected keywords that are important enough to boost
@Kevin_Lee_QED
27. Map Based Bid Boosts: How It Works
Custom map applied
$
$
$
Keywords selected
Maps data sync
AdWords
$
$
$
$
$
Geo campaigns created
• Customized map applied to selected keywords
• New geo structures built and keywords either campaign replicated into the
Google AdWords account or Bid Boost in the master account by geo
modifier
@Kevin_Lee_QED
28. Geo Boost: How It Works
Typical online campaign runs nationally
@Kevin_Lee_QED
29. Geo Boost: How It Works
Ex: A Map campaign targets many geo regions
Hundreds to thousands.
@Kevin_Lee_QED
30. Geo Maps: How It Works
• Demographics
• Household Income
• Family size
• Wealth
Maps can be built to suit many
needs
• Shopping cart size
• Conversion rate
• Client specific goals
@Kevin_Lee_QED
31. Map Definition Files: How It Works
• State (Region)
• DMA (Metro)
• City
• Zip Code*
Maps can be built to different
geographic sizes
@Kevin_Lee_QED
32. A Retailer Conversion Map
Hot conversion region
Cold conversion region
Conversion Rate by DMA
*Based on data previously tracked by Didit 2010-2013
@Kevin_Lee_QED
33. Maps: Household Income by DMA
Hot HHI region
Cold HHI region
For demonstration purposes only
@Kevin_Lee_QED
34. Maps: Wealth by DMA
Hot Wealth region
Cold Wealth region
For demonstration purposes only
@Kevin_Lee_QED
35. Geo Segemts - Benefits For Advertisers
• Weed out unqualified traffic (non-profitable clicks)
• Bid on keywords in areas that are more likely to buy
• Customize campaigns that appeal to people in a
specific geographical area
• Improve ROI performance and VOLUME on top
funnel/non-brand keywords
• Resuscitate a stale campaign
@Kevin_Lee_QED
36. Personalized Search
Big Data hits Google, again:
Already a fixed organic position across your target
audience doesn’t exist. Customization and
personalization happens based on:
1. Device
2. Geography
3. Login
4. ???
@Kevin_Lee_QED
37. Schema.org & Search
Schema.org opens up a whole new world of
information that the search engines can collect from
you (or from others about you)
1. Product physical properties
2. Product relationships
3. Specific product types
Still keyword based… But moving towards nirvana.
@Kevin_Lee_QED
38. Rich Snippet Benefits
Rich Snippets are Schema XML within the SERP.
Notice the relationships
between the data shown
are not keyword-based
but are a knowledge
graph.
@Kevin_Lee_QED
39. Rich Snippet Benefits
Rich Snippets have commerce benefits too.
Structured data is already
being used extensively within
paid search and that’s why it
is often more relevant than
organic. Yet Google has good
reasons to use metadata for
organic.
@Kevin_Lee_QED
40. Big Data in Earned/Shared - Social
Earned - shared social media have uses for big data,
in particular Twitter where data can be harvested
easily
• Research
• Prospect identification
• Influencer identification.
@Kevin_Lee_QED
41. Big Data fails in social if not “fun”
A simple mention, comment or “Like” is the beginning
of feedback and reinforcement for prospects and
customers.
However these forms of reinforcement are generally
labor intensive so consider:
• Delegation
• Filters
• Frequency
• Differences in platform (who sees what)
@Kevin_Lee_QED
42. Celebrities (Twitter and Traditional)
Followers are great but consider Klout, Kred
and upward influence (working on a scoring
mechanism for that).
@Kevin_Lee_QED
44. Celebrities, Learnings from #MillionDimes
Celebrity and # of followers (2.6 million for
example) are not necessarily correlated with
savvy Twitter behavior:
• What is wrong with this tweet?
@Kevin_Lee_QED
45. Old Search – New Search (Big Data)
Most of the old and today’s algos reputation converts
into rank/position.
Next Generation Search:
• Who do you know? Who does the searcher know,
like or follow?
• Who (individually) wrote the content matters.
• Who “liked” the content matters
• Who Tweeted / Promoted the content matters
@Kevin_Lee_QED
46. Strategies for New Search
Strategies for semantic search require a doubling
down on old school SEO Content and a focus on:
1) Diversity of content
2) User contributed (or generated) content
3) Promotion and sharing of content
4) Use of all assets for content creation
5) Tagging content effectively
6) Multi screen use of content
@Kevin_Lee_QED
47. Searching for Answers
Bing & Facebook have been active combining
Facebook’s social graph with search.
@Kevin_Lee_QED
48. Searching for Answers
FB graph search - who knows or likes what.
Suffers from lack of data and small friend lists.
@Kevin_Lee_QED
50. “FBX vsFacebook Advertising”
FBX lets you bring your own targeting via retargeting
and third party cookies.
Facebook also lets you bring your targeting via
“Custom Audience Targeting.” Both of these are more
likely to be successful than pure profile advertising.
What are your Facebook KPIs?
@Kevin_Lee_QED
51. Big Data, Search and Media Interaction
People don’t search in a vacuum. Media, PR and
marketing stimulate search behavior.
•When people get curious about a product they search.
•Brand searches spike most
• Quality score rises (long term impact of media)
• Category search may also increase if SOV is high
•Search becomes as much a metric as it is a media
This is why attribution models often neglect to quantify
the impact of failure to be there in search. Better to
leverage search.
Quality Score goes up when you advertise.
@Kevin_Lee_QED
52. Things That Drive Search Behavior
What stimulates search in your business?
•Online Display Media.
•Social Media (advertising and buzz)
• Email Marketing (CRM and acquisition/advertising)
•Direct Mail
• Offline Media
• Public Relations (On Good Morning America?)
Some of these are paid media, some are earned media.
How well can you control these?
@Kevin_Lee_QED
53. Social Stimulus
While search is likely to see the majority of the result of
other marketing stimulus, social media may also see a
spike as well.
•People may talk about you on Twitter
• People may like you (especially if that’s what you asked them
to do in the advertising or marketing message)
• Google Plus?? Will people add you to their circle?
• People may write about you in their blogs
@Kevin_Lee_QED
54. Back to Attribution Modeling?
Even better: Marketing Mix Modeling
User served display
Ad
User Searches
User served display
Ad
User Visits
Website
User Converts
USER
Marginal elasticity of every media option is key!!
e·las·tic·i·ty - ilàstíssətee
In economics, elasticity is the ratio of the percent change in one variable to the percent change in another variable. It is a tool for
measuring the responsiveness of a function to changes in parameters in a unit-less way.
Elasticity is one of the most important concepts in economic theory. In addition when understanding how to allocate
media dollars, nothing beats knowing the elasticity of the market in every direction.
Touch Point Analysis is only the tip of the iceberg.
@Kevin_Lee_QED
55. Long Term Interaction Test - Mini Mix Model
Have enough budget and time?
Split the country (or other geography)
•Raise spending on some or all media significantly in some areas
but keep control groups
•The bigger the spend the less time you’ll need for data
•Watch for increases in brand search impression counts
•Watch for increased CTR on brand and category ads
• Re-normalize the geographies
• Repeat switching the on and off geographies to be able to
validate both lift and decay.
Was there Brand lift, Quality Score lift?
@Kevin_Lee_QED
56. Big Data Personalization
(touchpoint with permission)
Check out this ad, dynamically addressing me by name.
Imagine how much more personalized it could get, with
permission & Big Data.
@Kevin_Lee_QED
57. Conclusion
• Harvest demand with SEM/SEO
• Create demand with Social, Display, early stage
SEO content, including video.
• Test things from greatest predicted impact down
• Work with external teams when appropriate
kevin@didit.com.
$200 adCenter coupon in my recent books *
* new US advertisers only
@Kevin_Lee_QED
Editor's Notes
- We also do Facebook and Twitter posts whenever possible. We found that Twitter hasn’t really worked as well, so we’re less motivated about those, but Facebook (particularly the cause pages) tends to be a great avenue to get the word out.
Explain importance of analyzing touch points and synergies between display & search to create demand (blizzard)