Credit Where Credit Is Due

Performance marketing has always been an industry built on data, but understanding the technology used to collect, analyze, and interpret that information as it is gathered in real-time has never been more important than it is today.

Connecting and interacting with consumers at every point of a non-linear product discovery and purchase path requires the agility to engage across multiple channels and the foresight to be present when they are ready to buy. However, the challenge of cross-platform attribution is not only about keeping pace with potential customers as they move from one medium to the next, it’s about understanding the dynamic between various media — the unique opportunities as well as the potential costs of cannibalization. 

In this first Technology Spotlight of 2018, performance marketers get very, well, technical about the latest advances in cross-platform attribution and discuss some of the growing pains that the industry will continue to address and work to improve throughout the coming months. 

TV Time
TV Time, an app with about 1 million daily active users, is just now
beginning to mine its expanding viewer behavior data and make it available
to networks, studios, and streaming video services — a possible boon for
sales attribution technology.

Latest and Greatest 

Cross-platform attribution has come a long way since the first omnichannel software systems began to look holistically at the variety of direct marketing channels running simultaneous campaigns.

“Most of the current cross-platform attribution solutions rely on probabilistic matching and modeling, whereby an element of chance is involved,” says Peter Koeppel, president of Dallas-based Koeppel Direct and a member of the Response Advisory Board. “With this approach, the utilization of IP addresses is common and provides a directional picture of cross-device usage. This is particularly useful for tracking behavior across multiple devices within the same location. The primary downside involves limitations in the ability to definitively track an individual across multiple IP locations (home, work, etc.) or an individual with multiple devices that may never originate from the same IP.”

However, he continues, new solutions are coming to market that use a deterministic approach, where, theoretically, data is known beforehand, which improves accuracy. 

“There are a few methods used to accomplish this type of match, and one of the most common is using an individual’s email address,” says Koeppel. “This is a solution still in its infancy. A benefit of probabilistic analysis is a larger footprint, but the tradeoff is that the degree of accuracy is dependent on the user’s trust in the data. A deterministic approach is inverse, with a smaller footprint, but a high degree of accuracy that leaves less to interpretation.”

Artificial intelligence (AI) and machine learning will continue to be a big part of the conversation as performance marketing solutions evolve.

“Google’s new platform, Google Attribution, uses these technologies to measure extremely large consumer datasets in real time,” says Fern Lee, CEO at THOR Associates in New York. “By tracking the entire consumer journey across various devices and platforms, the marketer is now able to make more informed and accurate decisions.”

The sheer volume of potential data and user behavior analytics to collect can be insurmountable. Algorithms help digest the information, but it’s most important for marketers to know what they’re looking for and to have top level goals already in mind.

“In recent years, as our multi-device world has gotten more and more complex, I think we have seen a massive advancement of mathematical solutions attempting to provide attribution across multiple touch points,” says Carol Hanley, chief revenue officer for Santa Monica-based startup TV Time, an app with about 1 million daily active users and a growing database of that viewer behavior. TV Time plans to sell that data to networks, studios, and streaming video services with Hanley, formerly a Nielsen and Arbitron (now Nielsen Audio) executive, leading the efforts.

“There are companies everywhere providing advanced data management platforms using AI to unlock a predictive technology that allows us to understand not just where a consumer was, but more importantly where the consumer is going,” Hanley continues. “These advanced analytical models have gotten more precise at attempting to provide real-time insights that enable a marketer to meet the consumer where they are going, rather than the traditional model of following them around five steps behind. These models are so advanced and would unlock profound insights if all of the data was there to populate them. They have the capability to go beyond ‘customer segmentation’ and predict individual behavior. That’s the holy grail. However, with as far as we have come mathematically, the reality is we still have massive walled garden environments where getting data out to allow for analysis in a broader context is quite challenging — and in some cases, impossible.”

Trust Your Instincts

Attribution will provide solid indicators and identify trends. It’s a good learning tool, explains Doug Garnett, CEO of Portland, Ore.-based Atomic Direct and a member of the Response Advisory Board, but it comes down to making smart judgment calls. As much insight as attribution can provide, it’s worth having some skepticism and remembering that it’s not going to tell you everything.

“The problem we have is that the underlying data is not yet solid,” says Garnett. “That’s one problem we’re fighting. Twenty years after TV was introduced, there was highly reliable third-party data that was thoroughly understood by all the advertisers that could be used to analyze the impact of TV. Here we are, 20 years into the web and we’re entirely dependent on the vendors themselves. It’s not independent data. We have to get it from Facebook. We have to get it from Google or whomever is supplying it, and it’s changing constantly. So as a developing medium, they’re way behind where TV was in becoming a mature industry.”

The difference, of course, is that the web is so massive. Some programmatic campaigns are buying on several hundred thousands of websites, he says, and there is no way to manage that kind of analysis, so “you’re put in the position of having to trust.”

Another challenge for performance marketers in this warp-speed environment is watching for and protecting against cannibalization between mediums. 

“One key strategy to evaluate cannibalization is to have awareness and measurement in place for all active marketing campaigns,” says Koeppel. “This enables a marketer to evaluate and execute strategies across all mediums at their optimal point. In order for such analysis to be effective, however, it is critical to have all aspects of a campaign under one centralized hub, such as an agency of record. This way the marketer’s best interests are not at cross-purposes with other vendors, such as competing agencies, where there might be competition for dollars or where the interpretation of data is tainted by self-interest.”

One of the most common pitfalls, according to Lee, a member of the Response Advisory Board, is incorrectly tagging a channel or platform with a sale and not attributing the appropriate acquisition costs.

“This can easily lead a marketer to make poor decisions rather than correct attribution that will lead to an optimized campaign,” says Lee. “What’s important to monitor are the device metrics alone and then together as a strategic indicator. For example, by looking at conversion analysis for mobile, it is important to gauge how many consumers called from a mobile device after visiting the site versus how many mobile consumers went to the site and didn’t call. Also, of note, many consumers use a bookmark to click and call in real time.”

Another pro tip for understanding the relationship between user activity on various devices?

“Monitoring and looking at how many leads convert and if they are profitable — or more or less profitable than the other lead sources — is a direct result of analyzing cross-device,” explains Lee. “When reviewing online data, the digital team can look at a baseline of how many leads it was receiving from people who went directly to the site prior to airing a DRTV media campaign and analyze leads versus conversion. The data should also include which browser was used (i.e. Chrome versus Safari), the analysis of returning visitor orders, as well as geographic location of the user.”

Understanding the limitations of the data is also one of the most importance factors in determining how to evaluate it.

“My experience is that the biggest challenge to measuring across platforms is that exposure is defined differently on each platform and even within each platform,” says Hanley. “The simplistic way is to assume every medium can be evaluated similarly — and in doing that you can cannibalize a certain medium because you suppress granular data in order to make it comparable to something else. The only way to avoid that today would be for the overall media industry to agree on a common definition of exposure. Minimally, the varying definitions of exposure by platform need to be taken into account and somehow made comparable.”

Google Attribution is one of the leading cross-device attribution technologies,
though it is still in beta. A paid version — Google Attribution 360 — includes
the linear TV platform, as well.

Making It Count

Though much has progressed in the way performance marketers collect and utilize data across various channels, the goals are ultimately the same: return on investment — and whatever the client’s other goals may be.

“With retail, we tend to start at the very broad high level,” says Garnett. “We want to look at gross retail numbers by day up against media spending because that’s staying at the level of what really matters. From there we’ll drill down, and you won’t see everything in that data, but you’ll see some things so you’ll follow a clue. This spending happened here and a week later there’s a bump — is that connected?”

And, he continues, it’s important to remember that not all activity will be neatly documented in a report. 

“One of the byproducts DRTV people always get is — because they’re on TV — they’re reaching a lot of people who aren’t going to buy today, and that’s a real big benefit to them. That develops their next wave of purchasers and the wave of purchasers after that. But you can’t measure exactly how big of an impact that had,” he says. 

After a DRTV airing, Lee says, the goal online is to generate an increased amount of conversions at a lower cost-per-acquisition (CPA).

“If the marketer observes the conversion rate or AOV (average order value) decreasing compared to its baseline, then there is most likely a disconnect between the DRTV creative and the website,” she adds. “This can be addressed by checking that the marketing message, offer, and creative assets are consistent across the channels. Additional metrics such as time on site, bounce rate, and frequency should be analyzed along with A/B testing.”

Everyone has their own method of using proprietary and third-party data to achieve specific goals, and while the industry recognizes the need for more consistent metrics from reliable and trusted sources, there are varying opinions on which one will fill that need.

“Among probabilistic cross-device solutions, I would say the leader is likely Tapad,” says Koeppel. “In terms of the deterministic approach, I would say Alphonso is at the forefront.”

Google Attribution is Lee’s top pick. 

“The program is currently in beta and relies on AI to offer cross-channel and multi-platform attribution reporting,” she says. “The system is able to collect data from ad airings via broadcast partners. TV attribution would require their paid version, called Google Attribution 360.”

Lee adds a note of caution: “The most unreliable way to track the success of a campaign is to look at ‘last-click’ attribution,” she says. “This is a widely used digital practice where the last touchpoint or channel a consumer uses is given full credit for the sale or conversion. Unfortunately, this model fails to analyze the prior steps that contributed to the sale. By not tracking the entire process, a marketer may leave out important information that the consumer learned during their journey.”

Hanley remains cautiously optimistic, erring on the side of using the data as a tool rather than a conclusion. 

“I don’t know of one unified system for cross-platform attribution because so many companies measure different parts of cross-platform marketing,” she says. “For example, a company that is great at broadcast (video) cross-platform measurement is not normally including print and/or billboard exposure for a total cross-platform unduplicated reach. A company needs quality, broad coverage across all media and scale to truly become the default. Until then, there’s lots of room in the proverbial measurement pot.”