In theory, Multi-Touch Attribution (MTA) sounds like a marketer’s dream come true.
Since digital media leaves tracks for each message or interaction, and since machine learning can find patterns in all that data, the idea is that MTA can determine the incremental impact of each ad, keyword phrase, video or downloaded white paper that led to a sale or other action.
That could be a godsend, given increasingly complex customer journeys, where would-be buyers move across online and offline touchpoints, often researching and making buying decisions before encountering a salesperson.
“If [MTA] was reliably accessible,” audience targeting provider Claritas SVP of Product Strategy Jeff Bickel said, “everyone would love the concept.” But it’s complex to understand and implement, he pointed out.
Accounting for the immeasurable
For one thing, he noted, MTA is a feasible concept as long as “everything is contained online.” But many campaigns and marketing interactions include offline touches, where it becomes complicated to capture and onboard that data with digital interactions.
Given this complexity versus the need to understand the value of marketing spend, he said, marketers are always evaluating whether their investment in a particular attribution solution is worthwhile, Bickel noted.
This leads to an ongoing question for MTA, he said: “Is the juice worth the squeeze?”
Offline data is not the only issue with MTA, of course. Brian Baumgart, co-founder and CEO of attribution provider Conversion Logic, points out that different industries — such as manufacturers — have different attribution requirements when you’re assessing individual impacts of touchpoints.
He also points to the issue of what he calls “immeasurability” — factors like existing brand equity that influence a buying decision. Apple has a built-in advantage in influencing a buying decision compared to a brand new maker of consumer products, for instance, simply because it is well known.
To accommodate this and other factors, Conversion Logic’s MTA implementation involves data models customized for each type of company and best-of-breed algorithms that employ advanced machine learning and Conversion Logic’s own identity graph, supplemented with graphs from Tapad, LiveRamp and Oracle.
Tony Marlow, CMO of data and marketing services firm Infogroup, finds MTA to be “critical” for his company. In working with attribution partners, he particularly likes the time-decay model, because it gives more credit to recent messaging and helps account for brand messages over the years, such as from car companies. There are several standard attribution models.
MTA and personalization
But he cautions that MTA is not “a silver bullet,” pointing to some additional complicating factors to consider.
Marketing personalization, for instance, aspires to deliver a message that is particular to one user at a given time in a given place, often optimized in real time.
“If you’re customizing creative,” Marlow said, “it throws down a challenge to MTA,” since the model needs to account for the impact of an ad for an ice cream parlor around the corner from a specific consumer on a hot July day, even though he just committed to losing weight. The finer the personalization, the less it would seem to fit into an attribution category.
But, for Bayer VP of Media Strategy and Platforms Josh Palau, even highly customized creative can be measured for its impact. “If [the attribution] covers creative,” he said, “we can see that a Watson [personalized] ad delivers twice what a display ad does.”
While MTA generated “a lot of excitement” when it first came out because it could assign value to touchpoints across both paid and earned channels, Claritas’ Bickel said, it requires massive yet granular data and is “pretty complex to understand.” These two reasons help explain why it is gaining traction mostly in big companies, he said, because they have the requisite data and trained staff.
But mid-range and smaller companies “aren’t fully embracing it,” he said, because first-, last- or first-and-last touches “suffice for a lot of places.” The path of least resistance, he added, is to go with “what they can touch and feel.”
First, last and brew-your-own
CallRail Director of Demand Generation Mark Sullivan agreed with Claritas’ Bickell, citing his sense was that there was “not a lot of understanding” or use of MTA in mid-market and smaller companies.
His own company, which offers phone analytics, has the challenge of obtaining and merging data from trade shows with broadcast and online marketing. While CallRail’s marketing includes Facebook, where it tracks leads via UTM tracking on links in ad campaigns, it is primarily focused on first-and-last touch.
In a typical path, he said, a user interacts with CallRail at a trade show, might later click on a Google ad and go to the company website, fills out a form, and, two weeks later, clicks on the company’s listing on review site G2 Crowd that leads to a signup for a free trial. But only the Trade Show and the G2 listing/free trial are counted.
Similarly, B2B customer data platform Radius uses first- and last-touch attribution, but with the addition of a credit to the team or team member that generated the opportunity.
Resources on Paid Media and Attribution
Engagement values
Merijn te Booij, CMO of call center provider Genesys, told me how his B2B company brewed its own sorta MTA solution.
At first, it went back and forth with first and last touch attribution, which didn’t provide the desired indication of ROI. Then, it assigned 40 percent credit to the first touch, 40 percent to the last one, and 20 percent to everything in the middle.
Finally, it decided to create engagement values, with each touch having 10 percent impact but additionally weighted with the company’s scoring of its own engagement value.
Using machine learning, Genesys processed its large data set of touches-versus-outcomes over the years, so that a comparative engagement value could be determined for each touch or sequence of touches. Similar sets of engagement journeys often led to similar outcomes, he said, even though there might be 30 touches in a journey. New sales to existing accounts, of course, followed a better known set of touches than, say, the engagement journey for a new account.
Similarly, domain extension provider .ME has also concocted its own multi-point blend that is “somewhere between looking at the numbers and trying to interpret what the numbers were saying,” according to CMO Natasa Djukanovic.
The company-specific complications for .ME is that many of its sales are through partners that also sell similar products and that jointly conduct campaigns.
For .ME, Google Ads campaigns lead to the best conversions, but they are only valued at 10 percent credit because they operate “more as a trend predictor” of several campaigns through other channels, she said. Google Analytics ends up being the most valuable tool, since it lets .ME determine the referral paths from partner websites.
‘A fool’s errand’: enter incrementality
Given its complexity and data requirements, however, some marketers see MTA as a dead-end.
“Multi-touch is a fool’s errand and it will ultimately die within the next few years,” ad platform Nanigans’ VP of Marketing Ryan Kelly told me via email.
“Because so many touchpoints with consumers are digital,” he said, the idea is that “they are therefore all trackable.”
Enough data would allow marketers to “build models to tell you exactly how much of an impact each touchpoint (paid or organic) was having,” he said, and the result would be a measurement of the resulting return-on-investment for each marketing component in a complicated customer journey.
The problem, Kelly said, is that “collecting the depth of data required to measure every touchpoint impacting a consumer’s buying decision is an insurmountable task in the real world.”
Besides, he adds, what the “smartest, most sophisticated brands” want to know is: “ ‘ How much net-new revenue am I generating from specific marketing efforts? ‘ “
Measuring the “incremental lift or incrementality of your campaigns will replace MTA,” he says, adding that incrementality has previously been ignored because it’s been difficult to implement.
Like a scientific experiment, incrementality in marketing runs a test group that receives the campaign or some variation of it, while a control group receives no campaign or an alternative one. This A/B test can then determine the net effect of the entire campaign, without having to assign values to, say, brand equity, hard-to-capture offline data or highly personalized messaging.
“Multi-touch attribution was a marketer’s dream,” Kelly says, “but it’s impossible to fully implement in reality.”
“Incrementality can not only be implemented, but actually solves the pain point of marketers trying to figure out: ‘How much business do I get when I do X activity relative to how much business do I get when I don’t do X activity?’”
This story first appeared on MarTech Today. For more on marketing technology, click here.
Comments