The fundamental challenge of attribution — understanding which marketing tactics and dollars are working — has only gotten gnarlier in the one hundred years since John Wanamaker’s lament that “half the money I spend on advertising is wasted; I just don’t know which half.”
The rise of digital marketing has offered brands the tantalizing prospect of end-to-end visibility into marketing impact. But the proliferation of channels, devices, tools, and vendors has combined with an increasingly non-linear customer journey to up the complexity of the attribution challenge.
Analytics leaders recognize the urgency of delivering real answers on attribution. Last year for the first time, digital accounted for more than 50% of total advertising spend in the US. And Gregg Archibald of Gen2 Advisors predicts that one of the major analytics trends for 2020 will be the increasing incorporation of AI into attribution approaches.
So how can analytics teams lead the way on attribution? Over the next three weeks, we dig into attribution to consider how it’s evolving and how the most successful teams are solving for industry-wide pain points.
It’s Complicated
The goal of attribution is to measure the impact of marketing spend on short- and long-term customer behavior. But the sheer number of channels and devices that a typical customer interacts with on the path to purchase — combined with the lack of a single standardized measurement framework — tends to create a massive reconciliation headache for marketing and analytics teams.
For example, a display vendor might claim credit on “view-through” conversions (i.e., they might take credit for any conversion within, say, a week after a user viewed their ad). At the same time, a search vendor might claim “click-through” conversion (i.e., they might take credit for any conversion in a session that resulted after a user clicked a certain search ad). In this case, both would be claiming full credit for the same conversion.
It’s not uncommon for brands to report $2-3 dollars of credit (or more) claimed by different marketing tools for every $1 of actual revenue.
The Standard Approaches
In order to provide a standardized framework for measurement and budget allocation, analytics teams that we work with tend to gravitate towards one of two approaches (or a hybrid combining elements of both):
Market mix modeling (or media mix modeling — MMM in either case) works by taking a top-down approach, using econometric statistical models to isolate the impact of changes in budget and marketing tactics on historical business outcomes. Most teams who endorse it view it as helpful for guiding budget allocation at the macro level, particularly around media that are not measurable with direct response data (like television).
Multi-touch attribution (MTA) looks at a huge range of converting and non-converting customer “paths” or journeys to algorithmically identify the differential impact of touchpoint A vs. touchpoint B. When used effectively, MTA has the potential to offer more granular refinements around digital marketing.
Both approaches have serious limitations that we’ll tackle in later blog posts (lack of visibility into key data sources, browser restrictions on cross-site tracking, privacy regulations like GDPR and CCPA, to name a few) but are battle-tested “tools of the trade” widely used across industries.
A Clean Customer Data Foundation
Regardless of which approach you are using for channel measurement and attribution — MTA, MMM, or some combination — it’s hard to overstate the impact of having a clean customer data foundation. This means an approach to identity resolution that is agile, accurate, and privacy-compliant (i.e, not fueled by hand-coded merge rules or outdated third party approaches) and provides a stable and persistent customer ID for every customer in your system.
We’ve observed a strong customer data foundation massively improving brands’ attribution efforts in several ways:
Knowing how many unique customers were actually acquired or retained can materially shift the view of channel effectiveness (particularly in MMM)
Being able to link converting customers’ path to purchase from one conversion to the next powers more accurate models that give differential channel weight depending on the purchase number (particularly in MTA). For example, a very upper-funnel channel like non-branded paid search plays a vastly different role in a customer’s first vs. fifth purchase.
The end goal is, of course, to give your customers better experiences so that they stay with your brand. Marketing analytics is a critical mid-point in making this happen so that you know what’s working and what isn’t. To reach the end goal, consider first things first: get the data foundation into fighting shape.
Learn how one retailer did it – and see the results – here.