• by Steve Smith , Staff Writer, November 28, 2014

Attribution remains the inscrutable but necessary element in the cross-platform march. Trying to attribute the impact of discrete marketing channels both apart and in tandem with others has always been a sojourn into weird math, even weirder algorithms and an increasingly large black box. Agencies that don’t want to build their own solutions are making their way as best they can among numerous vendors promising to crack the code.

To get a sense of how one agency is making its way through the attribution morass Gina Lee, SVP, Solutions at Boulder-based IMM walked us through how her company’s approach has evolved. Begun in 2007 as a performance agency, IMM has expanded since then into a more full-service agency. But among its early achievements was helping Guthy-Renker build its performance program for products like ProActive. “Back then the tools we had were all last-cookie-dropped got credit for the sale,” she recalls.

Then and now, IMM has its own in-house tracking to view its own cookie stack and determine from a user’s ID how many times they have touched that consumer and with what ad. “But it is pretty basic stuff,” she says.

Several years ago, IMM started looking for partners in attribution because its business was not about building these solutions itself. For a while they used only one of the key players in the market, Visual IQ, in part because “we were looking for stability.”

Also important was “the ability to ad serve alongside our ad server. Typically, an attribution platform will ingest your ad server data, so if you are using an Atlas or a Doubleclick, it will take the raw data fields and ingest them and do the analysis based on touchpoints and ideal paths to conversion, and put it up on a dashboard to take action against…which is fine and good. But we were also looking for information that maybe the data was lying to us. So we wanted a gut check on our own ad servers.”

IMM started comparing results, serving across several vendors to find varying results. One was picking up things others were not. “And if you are being paid on a conversion, then every conversion matters,” Lee says. “We are literally being paid on every pixel that fires.”

IMM works with Visual IQ now for everyday optimization, although it also uses Mediaplex. “What is great about attribution is that it is much more holistic. We know that search is very efficient at driving results, but search cannot work in a vacuum. If you go to any media mix company they will tell you to drive more search volume. But search is a demand-based media platform. We only serve ads when people search on them. It is not efficient to serve on every keyword in the universe. We need to have the other channels, including display and video.”

In the past year or so the increasing importance of cross-platform paths to purchase is where it gets even more interesting, she says. “It is not only video to display to search. We are now talking about mobile video to mobile search to mobile display to desktop search. That can be a real path to conversion. Behavior has changed.”

The real key in cross-platform tracking comes in understanding when, where and under what circumstances you already have seen a customer. The mobile search user clearly needs a different message and likely a different landing experience than the desktop searcher. “But if that same consumer comes back and you have seen them before, and you know you saw them on mobile search, and now they are on the desktop and clicked on a banner, you should change the messaging and understand they have already engaged with your brands. If they clicked, then you know you have done something correct.”

Ultimately, the goal of cross-platform advertising will be sequenced messaging that takes advantage of advanced tracking. Lee admits that mobile remains a choke point. She cites VisuaIQ’s recent partnership with TapAd as a step in the right direction, as it folds mobile Web tracking with in-app behaviors. It is all still new, however, and IMM is evaluating how to leverage desktop/mobile data for better campaigns.

But for now, they are getting close to sequential messaging by at least being able to test multiple brand messages at the upper funnel to see when the first exposure doesn’t get a response. “So we give you another message, not necessarily sequencing. But the idea is that we are testing them and that one of the three will get you to take an action. So then we funnel you down to a more aggressive appeal – like ‘do you know we have this and others don’t?’ Then we have three or five different messages, one of which might trigger action." Understanding what appeal worked to get you to take action becomes key to further messaging. “If you visit the site depending on how you came in – whether via a show statement or a sale or customer service message, then that is the message we continue to serve.” But the model also needs to be able to pivot with the consumer. “If you back up and click on a mid-funnel message – if you came out and are now intrigued by customer service – then we will change the path for you because that is more interesting.”

As much as the attribution tools and mechanization are helping her plan and optimize campaigns, Lee notes how the technology is emphasizing, not diminishing the human element. “I think attribution lets us be more creative around messaging. It is not ‘here’s the brief, come back with a concept and we let ir run for a quarter.’” Now it is important to think about the flexibility and adaptability of the concept as consumers react to it. “They are telling us what they want based on behaviors and time spent with us.”

The machinery has its limits in decision-making, she adds. “Machines can only go so far, and they help us tremendously. More than ever attribution is about the art of looking at the data. If it really were about this one number or this one KPI you wouldn’t need any of us and machines could do our job.”