As I recently wrote in A Tale of Attribution Woe (read that beginner-post first if the idea of attribution is fairly new to you), understanding attribution well (note, I didn’t say “figuring it out”) is an essential part of any digital marketing strategy. Unfortunately, it is also an evolving industry… which means there is still a lot of guesswork and change involved.
Attribution Ditch #1: Attribution Ignorance
When it comes to attribution, I believe there are two ditches that need to be avoided by marketers. The first ditch is the more obvious one: it is the ditch of attribution ignorance. That is, the difficulties of attribution are crucial to be aware of when setting budgets and assigning ROI properly, and it is no longer an excuse to ignore attribution. Whether it be cross-channel or cross-device, we need to get better at identifying how different channels impact our client sales. Going beyond a simplistic last click model in our understanding is essential.
In some ways, I would argue that this ditch has increasingly been called out and warned against successfully in our industry. There is still a long way to go, but attribution-awareness has been significantly increased from even a couple of years ago, and I find that even clients are hungry to unpack the puzzle of the attribution enigma in their accounts.
So we veer away from the ditch of attribution-ignorance… and head directly across the road into the ditch of attribution-arrogance.
Attribution Ditch #2: Attribution Arrogance
Whereas attribution-ignorance is undervaluing the knowledge that attribution can bring to a client account, attribution-arrogance is over-estimating the knowledge that can be gained. It looks at a simplistic model included in Google Analytics, assigns X % of value to each source, and confidently sends a report to the client, “thus hath the mines of mystery been plumbed, and thus shalt the budget be setteth.”
This is a ditch because it communicates to the client that attribution is simplistic, requiring only a specific formula (which BTW, the agency has clearly figured out that perfect formula </sarcasm>), in order for infallible ROI measuring awesomeness to be grasped.
Attribution’s Fatal Flaw
However, there is in my belief, a significant weakness to attribution that must be understood in order to keep our accounts out of either ditch and squarely in the center of the road. This center of the road, by the way, is always where the best marketers have proven themselves. It is the immaculate balance of both data and human-gut-intelligence-prowess. A good marketer uses data. A great marketer uses data to take action on what she believes to be true that has not yet been proven yet (and sometimes, can never be proven).
Regardless, let’s get back to the weakness of attribution. The glaring weakness of attribution is none other than our inability to accurately track human emotions.
What I mean by that, is that attribution will always be limited to the data it collects, when the actual decision made in a sale happens in the mind.
Allow me to illustrate this with one of my favorite characters (he was my favorite long before he was made popular by Benedict Cumberbatch!).
Sherlock Holmes is a master of deducing facts in order to solve a case, but not every fact and not every deduction holds equal value in the resolution of a case.
For instance, he may discover fibers on the floor that lead him down a mental path, and then he might interview a witness who lies about a key piece of evidence, and then this might cause him to visit the moor itself whereby he will put the finishing touches on the case.
Yes, attribution can answer the question: “which factual interactions led Sherlock Holmes to solve the case.” But attribution can NEVER properly weight those. I do realize never is a strong word, but I stand by it.
For instance, analytics of event facts and user behavior data cannot reveal the fact that it was the witness lying that caused Holmes the *most suspicion*, which led him to pursue the case more intentionally and thus visit the moor, leading to the resolution. Without Holmes (or, Watson for that matter, or really Sir Arthur Conan Doyle) actually telling us what went on in his head, we cannot know his intentions and how they were impacted by each interaction.
The weakness in using this as an example is that we can see into the head of Holmes, so we are brought into the decision. This is not the case with online buyers! While you can track user behavior on your site, and you can identify and fire various events to identify who did what, when on your site, you still can never actually know which channel caused the most “credit” for a sale in the mind of the user.
This is absolutely crucial because we analyze attribution data in a percentage model. This model doles out equal percentages of credit to the channels in the middle of a funnel, that model doles out 100% of the credit to the last channel to send traffic, and so on. However, these are doling out credit as percentages based solely on TIMING OF SESSIONS, and not on how we as humans actually make decisions… with emotion, with logic, with reason, with desire.
At the very least, one change that needs to happen immediately is less bold-faced ROI claims from attribution and more honest communication with clients into the actual state of things. Be less concerned with finding a 100% perfect attribution model, and more concerned with diving deep into a partnership built on trust that will allow you and the client to adapt over time as you continue to experiment and tweak their attribution model based upon source interactions over time.
So to close, as we think about attribution I’d like to warn us not to run from the one ditch into the other. Attribution is evolving in digital marketing (woohoo!), but our understanding of it needs to evolve as well.
We need to stop simply asking “how were the sources arranged in this transaction” (that’s a great place to start) and instead begin asking immediately after the first question, “what can I learn about my customer’s emotional orientation towards my brand in each channel.”
Also, frankly, we just need to be okay with not having attribution 100% figured out. We can’t know it perfectly. We can never know it perfectly. Take a deep breath, repeat that to yourself, and then get to work trying to get as close to perfect as possible in your client.
I would like to leave you with a resource for further study. After I wrote my Attribution and the Cows post on ZATO (linked above), I was contacted (and helpfully corrected) by Peter O’Neill of L3 Analytics. During our twitter conversation, he pointed my way to a presentation he gave way back in 2012 of the issues with attribution (I had already written this post before that conversation). I scrolled through the slides and thought they were some of the best work on attribution I’ve ever seen, so I wanted to share them with you here.