In many ways, the challenges around attribution have grown more complex as customer habits change, new marketing channels such as social media emerge, and even more data is added to the mix. The traditional marketing/sales funnel – from awareness to consideration to purchase – is anything but traditional in the current environment, making it increasingly difficult to pinpoint which campaigns influence buying decisions or other outcomes, either directly or indirectly.
In many ways, the attribution challenge is a riddle wrapped in a mystery inside an analytics program.
The ability to develop consistently accurate and reliable marketing metrics for attribution remains elusive for many marketers. A Microsoft/Atlas study found that between 93-95% of audience engagements with online advertising receive no credit when campaign ROI is measured. A Visual IQ study found that only 16% of major advertisers said they were using attribution models, while 47% said they were considering using attribution. Among agencies, only 18% said they were using attribution technology, with another 63% saying they planned to implement it.
As the challenges mount, it’s reasonable to ask whether full cross-channel attribution is an achievable goal.
The short answer is, maybe. There are many interactions between a brand and a customer or prospect that simply can’t be traced with complete certainty to a specific source. But that doesn’t mean there’s no room for improvement. The more steps marketers can take to understand how different channels influence conversions – leveraging emerging methodologies and technology, as well as some old-fashioned change management – the better positioned they will be to make informed decisions about marketing spend.
But first, a definition (or two). Attribution is a process that links marketing activities with an outcome (e.g., a purchase). Proper attribution is critical to ensure that you are funding the marketing programs that are most responsible for driving sales or other campaign objectives.
Forrester defines attribution in a digital context: “the practice of attributing credit to all marketing exposures that led to a website and subsequently resulted in a conversion event, rather than attributing all credit to the exposure immediately preceding the conversion.”
Simply put, attribution is about giving credit where it is due. Unfortunately, there is nothing simple about the online and offline paths that lead to a sale or other conversion. As the buying process becomes more complex, campaign attribution has become increasingly difficult to quantify.
Attribution 1.0: Current State
Although marketers intuitively know that many variables contribute to a conversion, the most popular attribution tools still tend assign most credit to the “last click.” This type of direct attribution credits the final interaction before a purchase or other desired outcome. In the digital world, the last click may stem from a keyword search, an ad impression, or an email click-through.
This last-click approach, however, ignores all of the preceding activities that also influence the outcome. A TV ad, related keyword searches, an online review, a social media recommendation all may “assist” a user’s path toward a conversion – much as a basketball point guard deftly delivers a pass to his teammate for an easy basket. In basketball, assists are tracked as a formal statistic. In marketing, assists are more difficult to discern.
It’s not for lack of trying. Most marketers realize that the last click doesn’t cut it anymore.
But they continue to rely on it, in part because it’s a tangible metric that is easy to track, easy to understand, and looks good on a spreadsheet. Another reason: A lack of standardized tools makes it difficult to properly assign credit to the other activities leading up to the last click. Attribution systems built on the last click don’t translate well across media.
In addition, they fail to take into account the non-linear paths that different buyers take through the marketing funnel. This new purchase model is “more iterative, circular and more about what the consumer is actually doing,” says David Edelman, head of McKinsey’s global digital marketing strategy. This updated sales funnel takes “a fundamentally different view of what’s going on in consumer behavior,” Edelman says. McKinsey even coined a new term for it: The Consumer Decision Journey.1
-linear “journey” creates havoc on last-click attribution models, requiring marketing teams to explore new methods for addressing cross channel attribution. At issue: the types of data they should be capturing and other steps they must take to develop amore granular – and accurate – view of marketing performance.
Companies are doing a better job of identifying what they’re trying to understand,” says John Lovett, a senior partner at Web Analytics Demystified who as a Forrester analyst wrote A Framework for Multi-campaign Attribution Measurement. “We’re seeing progress in terms of understanding customer behavior and the channels they’re using.”
This deeper understanding is leading to more advanced methods for assigning attribution.
Marketing attribution models are beginning to move away from last-click attribution, with a variety of approaches and methods emerging. Models such as “allocated attribution,” for example, are built on activities across digital exposures based on samples of cookie and clickstream data. This is an important step in establishing a method that accounts for direct and indirect effects of digital and social exposures within the context of all marketing and selling initiatives.
The key to this approach is identifying all touch points across a campaign that have some measure of influence over a conversion. This approach, also known as operational attribution, creates a detailed account of a random (or targeted) sample of website visitors – how they found your site, what they viewed, what ads they were exposed to, and so on, by capturing data across three key technologies: ad servers, website analytics and advertising analytics.
Once the sequence of activities is formed – a significant challenge in and of itself – an analyst can use weighting techniques, based on recency or other judgmentally-applied rules, to identify and assign a value to each of these touch points.
Another emerging attribution approach is media mix or algorithmic modeling. Algorithmic modeling involves the analysis of impressions, search data, email information and other digital sources to statistically correlate patterns and trends.
Algorithmic modeling can be used in conjunction with operation attribution to fine-tune the media mix.
These methods share a common goal: to provide attribution across the entire funnel. Most marketers will agree that full-funnel attribution is the only proper way to evaluate their efforts. What they don’t agree on: Which tools and techniques they should use to capture these insights.
Michael Brenner, senior director of integrated marketing for SAP, offers a 5-step solution for measuring what he calls “pipeline influence” for B2B marketers:2
Brenner notes that this approach “requires contact details for each marketing touch point, so website visits that do not gain registration, social interactions, ad impressions – none of these will be counted unless they gathered registration.”
Similarly, MediaPlex offers an attribution framework to help marketers identify the most relevant and most effective marketing mix variables. The framework requires first identifying the various “paths to conversion” – the sequence of events taken by a user on the way to a conversion event. Second, analysis of this data provides additional insight into “how the various marketing mix variables interact, where adjusting spending on the marketing mix variables can lead to more efficient conversion generation, and how marketers can create more conversions.”3
These are just two of many emerging methods of analyzing attribution data. There is no “silver bullet” framework or methodology that has captured the hearts of all CMOs. Technology solutions are just as plentiful; they come from full-service Web analytics vendors such as Coremetrics, Omniture and Google. Each offers different functionality, ranging from the types of data they manage to the modeling techniques they offer.
“I haven’t seen the perfect solution yet,” says Lovett. “Nothing can givey ou definitive, empirical evidence. But they will give knowledge and directional guidance that you wouldn’t otherwise get.”
Aligning the Business
For attribution to continue to evolve, marketers must get much more granular with their analytics and find ways to integrate offline and online metrics.
There are two steps in this evolution:
First, marketers need to push beyond the “easy” attribution methods and metrics, and take advantage of the depth of data they are collecting to unlock the true insights embedded within. Sampling techniques, for example, inevitably over- or under-state true attribution since it is virtually impossible to sample without error, and that error gets magnified many times over when millions of customers and billions of engagement points are involved. It’s sort of like the children’s game of “telephone” where the real message gets distorted terribly by the time it comes out the back end. Add “judgmental” weighting on top of that and you have a prescription for compounding layers of human error.
The new generation of “big data” analytics provides the means of processing many terabytes of clickstream and ad–server data very cost effectively to accurately assign “assists” where assists are actually observed. They actually count each and every touch point in the customer or prospect online engagement and then assign assists accordingly. Each possible path way is measured by relative frequency of observation, highlighting those found most efficient and productive.
Solutions Offering Insight into Attribution are Available
“Today, there’s no longer any need for making judgments about attribution,” says Wes Nichols of MarketShare. “The technology required to capture, store, and count the frequency of online engagement and exposure is available today, and at reasonable cost. By combining smart use of cloud computing and advanced query languages, we can quickly sift through enormous amounts of data to find the real pathways to conversion.”
Second, these “big data” discoveries need to be incorporated back into a comprehensive analytical framework which measures cross-tactical attribution across the full spectrum of online and offline elements. It’s important to derive accurate digital attribution, but pretty useless if you can’t put it into the larger context of all other offline tactics. The analytics involved are not trivial, but neither are they rare any more.
Technology and methodologies, however, are just two pieces of the puzzle for advanced attribution. Most CMOs have learned by now that technology is most effective when used as an enabler, not as a driver of strategic initiatives. With that in mind, for attribution capabilities to succeed in a marketing organization, CMOs must address two additional elements:
The data being collected for attribution frameworks, and the culture of both the marketing organization and the business at large. Both loom as significant challenges.
Data: Large organizations have invested millions in master data management or customer data integration initiatives in an effort to capture the elusive” single view of the customer.” But most companies continue to struggle with the most effective way to integrate multiple data sources – a critical step on the road to the next generation of attribution capability. Social media only adds to the challenge. What’s the best way, for example, to integrate Twitter data streams into salesforce.com? More importantly, is it worth even trying? CMOs must match their marketing measurement and attribution objectives with the many internal and external data sources that marketers collect to ensure that the right data is not simply being captured, but that it is accessible by attribution tools.
Even if the data itself isn’t centralized, the analytics should be. Mediaplex explains why: “Even if marketers are using various, independent solutions for managing paid search or email, for example, those marketing channels should still use a unified analytics platform for tracking clicks and impressions. Moving to a single, comprehensive platform allows marketers to accurately track conversions and view consumer interaction data from ALL online channels and sites, as well as establish the baseline for how they will define the data for any future analysis.” 4
Culture: Perhaps the most imposing barrier is the need to bring the various marketing silos together under a shared vision of marketing optimization. There are plenty of politics involved in marketing resource allocation with every group competing for limited budget dollars. Don’t think for a minute that you won’t get push back once your attribution model starts taking credit away from one group and assigning it to another.
“Changing the way firms measure and allocate something as integral as marketing spend often requires a rebalancing of internal politics, which can be much harder to assess” than attributes such as the availability of data, Forrester analyst Fatemeh Khatibloo.5
Advanced attribution requires marketers to understand how all parts of the ecosystem in which customers engage with a brand influence behaviors.
This will include elements marketers can control as well as those they can’t: the state of the economy, unemployment, social sentiment, etc. Attribution 3.0 is about connecting the dots between online and offline marketing campaigns and online and offline revenue.
drive sustainable changes to processes and the organizational mindset.
While no two models will be the same, organizations that can develop an attribution framework that delivers the best mix of consistency and rigor will be well positioned to maximize their resource allocation and improve the return on their marketing investments.
Pat LaPointe is executive vice president of MarketShare (www.marketshare.com) and editor-in-chief at MarketingNPV (www.marketingnpv.com)