“In God we trust; all others must bring data.” – William Edwards Deming
If you’re a marketer today, this quotation might be scribbled on a sticky note stuck to your monitor, or framed on your desk, or set as your screensaver. As painful as it sometimes is to measure your work with data—that cold unfeeling judge—marketers today can’t escape it. If you control a marketing budget, you understandably want to know how every dollar performed so you can improve efficiency in the future. Say every dollar you spent on search engine marketing drove ten dollars in revenue, while every dollar spent on display returned five dollars. You might, on that basis, shuffle some budget away from display and into search in hopes of driving more revenue per dollar spent in total.
Attribution is how marketers know what their Return on Investment is for each marketing channel. In theory, attribution works like this: take all the revenue that your marketing influenced (i.e. incremental revenue; you don’t want to give your marketing credit for purchases that would have happened anyway) and spread it out fairly according to how much each channel influenced those sales. Now, as usual, this is much easier said than done. In practice it’s quite difficult to determine a) how much revenue is incremental, b) which marketing channels a customer touched on the way to a purchase, and c) how much credit a given marketing channel deserves based on its contribution to the sale.
The simplest version of attribution found in the wild is known as “last click”. Last click attribution gives all the credit for a sale to the last marketing channel the customer interacted with prior to the purchase. Using last click eliminates the complexities around (b) and (c) above: it ignores all the touches in the customer journey except the last one and assigns 100% of the credit to that touch. But it’s suspected that last click doesn’t provide a very good picture of your marketing ROI, which is the whole reason you’re doing attribution in the first place.
So how do we get better intelligence about our marketing budget? Let’s assume that we can resolve concern (a) above, and know which revenue is incremental (note: this is not a problem to be tossed off with an assumption, but requires more in-depth analysis than we both have time for here). Now we just need to solve the last two. Concern (c) is a theoretical problem—you can solve it on a whiteboard—and therefore doesn’t have the same practical roadblocks as (b) does. It’s probably best if you just read Avinash Kaushik’s post about attribution models if you’re concerned with (c). That leaves us with one problem to address here: (b).
Most marketers today have no way of knowing all of the marketing channels that led a customer to a purchase. This is because they track their users’ behavior with cookies, which makes it impossible to understand activity that spans across multiple devices. If I click on a display ad on my phone that takes me to your site, and then I click on an adwords ad on my laptop a day later and purchase, you won’t see the mobile channel at all in my journey. Therefore, your attribution won’t assign your mobile campaign the credit it is due. Brutal, right?
In general, because of the way that you probably track your customers today, your attribution is going to cause you to undervalue the marketing channels that don’t directly lead to conversions, or that live on devices that shoppers don’t usually convert on (i.e. mobile). Because you can’t link a user’s behavior on their phone to their purchase on another device, you’re missing a lot.
Attribution can give you novel insights into your marketing, or it can prove what you already intuitively know. Either way, you need the data to backup your hard work so you can keep getting better—or stay at the top of your game. If you don’t have a cross-device and cross-channel view of your customers, your attribution is leading you astray; it’s kind of fibbing. The good news is that better insight into your customer’s behavior across all their devices could get your attribution in shape without you having to change your attribution provider. Your problem today is with your underlying data, not your tools.
If you’re reading this, I’m guessing you’re not God. So you must bring data, in line with the credo that opened this post. But before you do, do yourself and your organization a favor and make sure that data is worth bringing.