Outcomes, Incrementality and Advocado's Approach

By Harold S Geller, VP of Strategic Development, Advocado - An agency veteran driving industry relationships and spearheading new strategic partnerships

Attribution—the science of quantifying cause and effect, has been co-opted by marketers and transformed into the science of allocating credit to exposures for driving sales or other outcomes—is a central, yet complicated and elusive topic facing the television industry. It is exciting and promising, but also confusing. There have been a vast array of approaches (some more useful than others), and many disparate datasets that fuel the analyses.

While many vendors like to tout their AI this and ML that, all of this technology is useless without:

1. A clear understanding of what is being measured, and
2. Clean, verified and reliable data from which the analysis is being conducted.
3. Ability to understand, trust and leverage the output of the attribution system

Successfully navigating all of these issues is essential because the promise of properly quantifying the value TV advertising has for marketers is enormous: the ability to gain context on how a TV ad campaign influenced consumer behavior (i.e. changes in specific search activity) or identifying the incremental lift in web traffic or sales driven by television advertising at a very fine level and at a pace quick enough to enable the tactical optimization of campaigns, mid-flight. The ultimate benefit, though, is moving beyond simply measuring impact or ROI in a rear view mirror to actively harvesting consumer intent, managing ROI across media, and measuring meaningful outcomes.

Advocado’s approach to attribution starts with the collection of proprietary TRUSTED, granular, first party data ad occurrence data which is used to audit media buys by the largest agencies and brands in the world:

• It starts with a serial number. Every piece of Ad content is uniquely watermarked
  • Critical standardized metadata regarding the spot is captured at the time of encoding, including customer name, product name, commercial title, and commercial length
  • Hundreds of custom data features, like campaign type associated with the campaign and commercial, are applied to each asset based on customer requirements.
  • Advocado monitors broadcast activity 24/7 and transmits the commercial detection data to our cloud in real-time.
    • • 210 U.S. DMAs and 121 Countries globally
  • Note: unlike ACR fingerprinting techniques that “think” they know which ad they have detected, by using the Advocado watermark, we know “exactly” which ad is detected - even if the ad is in an inset window within programming or if an advertisers creates 10 version of the same ad and only changes the call to action graphics in the last 5 seconds of play.
    

• Customer’s Website receives our first party web tracking tag to bring granularity back to web activity data
  • We can identify every unique session (not just counts by the minute)
  • Follow page views & time on site
  • Track conversions. Page paths, session duration, shopping carts, single-page checkout, etc.
  • Connect conversion with originating session, even if it occurs minutes, hours aor days after the initial session.
  • Collect tracking codes like UTMs, referring sources and even things like the Google Click ID that comes across if it's a paid search visitor.
    • Our API interface to Google, allows us to extract the literal keyword matches and exposure to online behaviors like website engagement, versus the regular behavior, which helps understand what your specific ads are doing to drive consumer questions.

Our approach is radically different from other providers:

• For creative executions, other providers, begin with creative fingerprint database matching that uses fuzzy logic matching algorithms that are error prone, and imprecise when it comes to creative nuances like customized dealer tags, calls to action, commercial length, and lack the ability to do answer basic questions like did the entire ad run in a market, or was it cut short?
• Other providers depend on syndicated web analytics platforms like Google Analytics that average and aggregate visits and traffic by the minute, which causes the loss of precision related to the visitor data.

The advantage of our data collection approach is that there is no guesswork, no assumptions, no modeled data, just granular data with standardized metadata, and custom features that can be used as inputs to a variety of critical comparisons that underpin questions about the impact of television advertising.

The other thing that sets Advocado apart from others is that we embrace both rules-based (directional) attribution and neural network based (incremental impact) measurement in the MicroMoment^Ⓡ, around the time of a linear ad exposure. By utilizing these two complementary methods of attribution, we can extract specific data, like actual paid search keyword activity inside and outside of the MicroMoment^Ⓡ and provide an unbiased measurement of impact - other call this outcomes measurement, incrementality, etc. - whatever you call it, you need to start with an unbiased approach and not the objective of “proving” one media is more (or less) worthy of receiving credit for an outcome.

While others use “Hungry, Hungry Hippo” black box style attribution, where credit is claimed for a consumer engagement as they try to prove a point, Advocado developed (and patented) its own approach to incremental impact measurement and then partnered with Amazon’s SageMaker DeepAR forecasting algorithm, to build out a state of the art, completely transparent, glass box model that makes it possible to understand and access the underlying prediction engine and its data, enabling you to develop trust, derive deeper insights, make detailed plans, and even run complex simulations aimed at improving lift.

Our approach is a counterfactual, deep learning neural net that doesn't try to “claim” credit for the TV advertising activity, but rather builds a predicted forecast about what would have happened if a targeted stimulus was eliminated - in this case TV advertising.

Advertising measurement is no longer ONLY about demographic Gross Rating Points, Reach and Impressions, or more generally the “upper marketing funnel metrics” of awareness, it is about mid to lower funnel metrics of action, attribution, and sales lift.

Advocado’s counterfactual neural network is built for known and unknown factors and isolates the lift based on the features of the data provided by the customer, that allows the surfacing, and explanation of features to have the most significant positive or negative impact on that lift.

In addition to using the metadata, and features provided by the customer, Advocado use raw weather station data to overlay weather conditions, closed caption streams from content adjacent to the linear ad exposure, and fingerprint based airings of the customer’s competitive set in our neural net to enhance the explanations of the lift caused by the linear ad exposure.

In addition, using the client provided metadata (called data features), “what if scenarios” can be run to model the different results, and how they impacted each individual event that we're attributing.

At Advocado, “Data is our Core,” not Linear TV or any other specific media; so we're not trying toprove Linear TV works or doesn't work, we believe that by starting with better, unbiased data and building technology without an agenda we will uncover unique insights and identify what really happened. As a result, our customers are able to make more informed decisions about their Advertising spend to derive better outcomes and trust the data enough to embrace next-level strategies like cross channel campaign automation.