2. data sciences: taking the guesswork out of targeting
Think about it. The billions of advertising impressions being served today contain a wealth of information about
the intersection of user, content and message. What was the product? Who was the consumer? Where did the
ad appear? When? How did the user react? All this exhaust from advertising placements provides a fascinating
stream of data that, if unlocked, can help marketers better understand and optimize their commercial messaging.
More than context or geography or behavioral attributes alone, it is the analysis and manipulation of this
information trail, in combination with historical and third party provided data, which is leading to the greatest
advances in campaign evaluation, targeting and optimization. Data sciences is leading the audience buying
revolution today, focusing on making sense of complex data analysis, audience modeling and machine learning.
audience modeling
While using data to target audiences is generally understood to be a far more efficient method of reaching a
desired audience over buying sites, what many advertisers don’t realize is that the pool of available, reliable
online audience data is generally small. Collective’s own analysis illustrates how the data richest 30 percent of
impressions have consumer attributes from several data providers, while the data poorest 30 percent have no
data at all. Therefore, sophisticated mathematical models have to be created to “fill in” the missing data on
audiences with less available, actionable data associated with them. This approach leads to greatly expanded
audience pools, with superior lifts in campaign performance. While audience segmentation and data modeling
isn’t a new concept in marketing (direct marketers have been doing this for decades), the techniques that are being
applied to online advertising are
providing marketers with
greater precision with near
real-time optimization.
Take the example of an
advertiser trying to reach
potential auto buyers. One
tactic might be to identify
consumers who have recently
visited Kelley Blue Book, or
people who have consumed
a lot of auto content
cars recently. Layer on
demographics and household
income data and you are
off to the races, right? The
reality is that only a fraction
Data sparsity image from Collective’s Data Sciences Team illustrates of that combined audience
a common myth about audience data, and the need for modeling. is likely to constitute explicit
auto intenders.
3. Intelligent Audience Buying
Using data to connect the who and where online
58%
The key is to model the user. If you can identify auto buyers,
particularly through 1st party data, you can look across all their
data attributes, look for commonality and greatly expand your
universe of potential customers. In this scenario you are less
reliant on individual data points because the model population
is modal, and that’s far more effective than trying to target a fixed of U.S. Interactive Marketing
data set.
Professionals say they
would increase their
precision versus accuracy online spend if better ad
The promise of real-time bidding and the sophisticated
deployment of audience data have led the path from media
targeting was available.
buying of content sites to the buying of audiences. Clearly, data Source: Forrester Research, 2010
is key. But too often it has diverted attention from where it needs
to be, which is accurately engaging the desired audience.
The distinction between precision and accuracy is often lost on advertisers and agencies. In our striving for
evermore precise targeting data points, accuracy is often lost and we miss the forest for the trees.
Developing a broader, more complete sense of your audience is fundamental to successfully deploying data
driven audience buying. Fortunately, data analytics can provide the insights marketers need to uncover the
habits and attitudes of their online audience.
At Collective, we’re developing proprietary audience segments using a methodology that provides both quality
and scale. Instead of focusing on the most active cookies, we look for an even distribution of cookies across our
audience cloud that conforms to the data attributes we are seeking. In this way, our audience is less likely to be
over-targeted and represents a more realistic portrait of the consumer.
HIGH SCALE
• Registration
Collective Segments
90%
• Context Inferred
of data falls into
these categories
• Verified user-supplied
LOW SCALE
• POS
LOW QUALITY HIGH QUALITY
4. what type of audience buying data
is available?
Audience data comes from a variety of sources and can be loosely
divided into the following categories:
Audience Verification:
Trust, but Verify Contextual -Many contextual segments are driven by keywords,
which can be risky for advertisers since contextual targeting can
sometimes present a misinterpreted meaning of a keyword.
There are many data providers and We’ve all seen notorious examples of bad contextual targeting,
advertising intermediaries selling for example Barry Bonds vs. Junk Bonds. Collective’s Personifi®
target audiences at scale, but how semantic classification looks for the meaning of the content,
do you know that you are actually
rather than looking simply for keywords.
getting the audience you are
paying for? Audience verification
is one of the most significant new Demographic - Used by most advertisers, this usually includes
developments in audience buying age, gender, income, or education. This type of data is usually
which helps to answer that question. gleaned from website registrations and through large offline data
Just as companies like DoubleVerify aggregators.
verify that ad creative is running in
the right environment, audience Lifestyle – Based on a group of users’ likelihood to behave a
verification seeks to measure how certain way, this modeled data places broad sets of people into
accurately ads are being targeted to defined segmentations, as one would see in a Neilsen Prizm seg-
the desired audience. ment like “busy mom” or “health nut.”
Collective is on the forefront of
In-Market – This type of data reflects the lower end of the sales
audience verification, regularly
auditing our audience data against funnel as it reflects users who have searched a specific topic or
known audience panels such as have visited an In-Market vertical site, like auto or travel. Sub cat-
comScore in order to measure the egories of In-Market data may include Search (based on a user’s
accuracy of our targeting. Our tests search history – right down to a specific brand or model number),
have shown a great disparity in the and Re-targeting, in which a visit to a webpage or exposure to
accuracy of available data from third- specific advertisement is used to infer the user’s interest.
party suppliers, so both marketers
and their advertising partners need Purchaser– Data that reflects ownership or consumption of a
to be cognizant of their audience product. This is often used by auto marketers who are
data sources. interested in reaching people who have purchased a specific
auto brand. It is also a great tool for conquesting.
20%
Data Type Usage Contextual Total
Source: AMP®, 2011 33% In-Market Total
Demographic Total
22%
Lifestyle Total
Purchaser Total
12% 13%
5. Intelligent Audience Buying
Using data to connect the who and where online
4 useful tips for audience buying
1
learn about your online audience first
Sure, it sounds obvious, but we can’t tell you how many clients
have come to us with definitive, inflexible definitions of their online
audience. Often this audience definition was drawn from offline
research or traditional media buying assumptions. But online
behavior doesn’t always align with offline customer data, and the
best way to learn about your true target audience is to gather
real-time online performance metrics by appending advertising
performance analytics with third-party consumer data.
2
it’s the “who” and the “where”
It’s important to understand that the audience is modal, switching state of
mind with each click of the mouse. The same mom you’re targeting when
she is on a casual gaming site is in a very different state of mind than when
she is looking up recipes. Matching the right person with the appropriate
context is essential for brands. This is one area where ad exchanges create
a potential danger for brand-conscious advertisers. For a direct response
advertiser, it might be fine to reach an intender anywhere they are. But for
those seeking to influence brands, the quality and type of content in which
the ad sits matters.
3
intelligent creative and the magic of DCO
Data can be used for far more than just audience targeting. In fact,
when applied correctly, data can be a powerful tool for optimizing
your advertising. You can enhance the relevance of your creative
and drive better results for your brand messaging with the appli-
cation of dynamic creative optimization (DCO). By setting up an
ad template to dynamically self-populate with optimized creative
components, including audience location, demographic and psy-
chographic data, you can create custom ads and tailor them for
individual customers in real-time. This is an especially great tool for
retailers who can tailor the products they’re advertising in digital
catalogues to precisely targeted viewers.
6. 4
don’t chase clicks
Over the years there have been a number of studies that have
examined the fallacy of using CTR as a measure of campaign
success and the nature of the clicker’s themselves. In new analysis,
Collective’s Data Sciences team examined the advertising
interactions of over 100 million anonymous user profiles and over
one billion advertising impressions to find that:
• 99% of stable cookies examined never click on an ad
• Nearly 20% of ads that received any click activity received
multiple clicks within the same
impression, suggesting that these clicks were unintentional
• Intentional clickers are lower income consumers
• Clickers demonstrate lower post impression actions and
brand lift
This study should serve to caution marketers that relying on
CTR means being comfortable targeting low income, older,
technologically less sophisticated consumers and recognize that
most of the click ‘leads’ will go nowhere, as they were generated
by unintentional clicks or will not result in a post impression action
or brand lift.
5
don’t over-target
The industry’s great ability to aggregate data and narrowly focus
campaigns sometimes leads to the desire for a mythical, exact
audience. You know, the right-handed green-Volvo-driving, health-
conscious grandmas in the market for new nutritional supplements.
And we could find them. All three of them.
So unless the product is extremely specific, cast a wider net. By
casting a wider net, you can find broader audiences that perform
-- ones that you might not know exist and that are often much more
scalable.
One of the mental pitfalls that search ads have created among
marketers is the over-emphasis on being down the sales funnel. But
we suggest that we not use display advertising to target customer
who’ve basically already made up their purchase decision. Instead,
create awareness and desire. By targeting the right type of users
at a higher level in the funnel, you help those users make their way
down the funnel, which in turn helps drive your search campaigns.
7. about collective
Collective provides comprehensive display advertising solutions for
advertisers and publishers. Our AMP® Data and Media Management
platform leverages the most trusted third party data, proprietary audience
modeling and ad effectiveness metrics to simplify audience buying and
selling across display, video and social media. AMP powers the ad
businesses of over 50 leading media brands, including our flagship products,
Collective Display and Collective Video®. Collective is headquartered in
New York with offices in Atlanta, Boston, Chicago, Dallas, Detroit, Los
Angeles, San Francisco, London and Bangalore. Collective’s investors
include Accel Partners®, Greycroft Partners and iNovia Capital.
Visit us at collective.com
contact
press contact:
Laura Colona
Director of Communications
lcolona@collective.com
sales contact:
contactus@collective.com
®
99 Park Avenue, 5th Floor
New York, NY 10016
888-460-9513