Attribution is rarely straightforward. While some of your sources have direct-response tracking, other sources will rely heavily on methods of matchback attribution. A robust methodology will define attribution in order of likelihood that a particular conversion originated with a specific campaign. Moving "down" the scale of likelihood is synonymous with attribution decay.
Attribution decay is similar to the way in which a decrepit building stands up while falling apart. When you look at the roof you'll find it still standing, but identifying which supporting beams provide structural integrity is an imprecise science. In the same way, cohort attribution will exhibit greater certainty in localized areas, and severe uncertainty in others.
An attribution decay can be as simple as a waterfall of Yes/No rules that define the attribution funnel, or you can make it as complicated as a statistics based methodology using time decay for observable in-funnel actions. Whichever you choose, use specific data points relevant to your business model and industry in order to define the most logical sequence of attribution points.
Define "perfectly" attributable data points
The reason attribution has a decay is because your analysis works down a scale of "attribution certainty".
The starting point for answering your toughest attribution concerns is to recognize that you already have trustworthy attribution on many of your marketing campaigns. These campaigns will tend to be digitally skewed such as SEM, however, some of your less traditional offline campaigns (such as in-store warranty upsells) will likewise have very good attribution.
Fill attribution gaps with the perfectly attributable
Once you define "perfectly" attributable campaigns, dissect the opposite end of the spectrum where campaigns are prone to misattribution. For these sources, such as print media or TV, build scenarios in which you can capture partial data points for reattribution.
Here are a few questions you should ask when defining campaign tracking rules from the angle of attribution:
- Can you use voucher codes?
- Can you use vanity URLs?
- Is there likely to be an increase in brand related SEM, organic search, or direct traffic?
- Does any combination of device type, time of day, or location give you attribution?
- Can you further reattribute untracked revenue?
Vouchers and vanity URLs give you the first two steps for filling in attribution gaps. Because brand and direct traffic are indicators of primed intent there is likelihood the traffic free-rides momentum from poorly attributable campaigns. Setting up rules around revenue reallocation from these traffic sources based on data points of location, time of day, and device type can further give you reason to attribute revenue to a particular campaign. As long as the campaign shows basic signs of generating revenue from these tools, you can likewise reattribute revenue as per the methodology previously discussed.
Basic attribution decay example
You have an ad across your local public transit. Here's how you could define your campaign's attribution decay:
- First allocate any campaign specific voucher revenue
- Allocate any direct revenue from vanity URLs
- Proportionally reallocate branded SEM, organic search, and direct revenue based on campaign spend
- You should only reallocate revenue generated on or after the campaign within your constrained time period
- You could decide to further constrain this reallocation only to traffic generated in the metro area of your campaign
- You could also decide to constrain this reallocation only to traffic from mobile devices
Recap: steps for defining attribution decay
Defining your business-specific attribution decay is an unavoidable part of any relationship between analytics and marketing. Collaborate on deciding which trackable elements are most important to your marketing optimization efforts - your ability to scale marketing spend in the future depends on your ability to attribute revenue to your campaigns with a certain degree of certainty.
- Define all available paid and unpaid marketing mediums
- Define which mediums give you "perfect" attribution
- Reverse engineer data touch points of your most unattributable mediums
- Repurpose attribution from perfect campaigns to fill in gaps based on a series of rules
- Finally, reattribute proportional revenue from primed traffic sources
Be vigilant, allow for exceptions
Nothing ever is perfect. Once you've set up your attribution rules you should actively second-guess yourself as there may be external market forces that are skewing your attribution.
My favorite example is that of voucher code "leakage" onto voucher and coupon affiliate sites, where a tracking mechanism intended for one medium makes it onto an unintended one. These things happen, it's unlikely to create a life or death scenario... just create a strategy for cleaning up your attribution.
Finally, define rules around mixed attribution, where perhaps you should reallocate 100% of revenue to a voucher code or vanity URL for traffic that appears to have a perfectly attributable touchpoint at a later point in time.