In anticipation of his upcoming keynote presentation at the eMetrics Summit Boston, “Optimizing Your Process for Being Data-Driven,” we asked Tim Wilson, Partner at Web Analytics Demystified, a few questions about his work in digital analytics.
Q. Where does digital analytics sit in your organization and how does it interface with your business units?
Well, I’m a consultant, so I sit in the only group we have — the consulting business unit! For all of my clients, though, digital analytics sits in the marketing organization — on the business side of the house.
Q. What’s the most valuable thing you’ve learned so far in 2014?
I’d like to say I’ve learned patience, but that’s never-ending self-tutelage. I think what I can most definitively put a pin in when it comes to what I’ve learned has probably really been more the culmination of the past 4-5 years, but I’ve really been validating it over the past 1-2. And, that’s the need for simplicity and clarity when communicating with business stakeholders about analytics. This occurs on multiple levels. The no-brainers are things like, “Don’t talk about s.props with the CMO.” But, there’s a more nuanced level where things really get exciting that has to do with the overall language and approach we, as analysts, use with our business counterparts. I no longer ask marketers, “What are your KPIs?” or “What are your success metrics?” or even “What are your hypotheses?” Even though they know the meaning of those terms, that is still analytics-speak. I may be looking for the answers to those questions, but I am constantly seeking out ways to ask them in even more marketer-friendly ways: “Why are we doing this campaign?” “What judgment calls did you make when planning this campaign that you were the most uncomfortable with?”
The same goes for when I’m presenting results of an analysis. There are scads of analysts out there begging analysts to do a better job of communication results — Ian Lurie at Portent and Lea Pica at Prudential Financial both make this case well (and have done so at eMetrics!) — but the overwhelming majority of our profession still doesn’t get it. I’ve learned that it always is worth the time to make what I’m presenting (or even sending in the body of an email) as clear and clean as possible! I had a client a few weeks ago comment, “Your Excel stuff always looks so nice.” We’re talking Excel, and, in this case, almost none of the files that she looks at from me have any images, logos, or even charts! But, I obsess about organizing, formatting (including conditional formatting) everything so that it is clear to someone (a non-analyst) who has never seen it before. If the client sees that as “nice,” I’ll take it! The fact is, it is clear and clean and professional…which makes it efficient for the recipient to digest!
Q. What’s the latest analytics method/process/tool set that you have implemented and what advice would you give others?
I’m going to cheat on this one and go with two:
First is super-tactical, and it’s a little Excel thing: conditional formatting using horizontal bars. To what I just said on the last question, it’s worth it to make every piece of information as clear and consumable as possible. I love-love-love taking a table, adding a column next to each column of a metric, setting the values in the new column equal to the metric, and then putting horizontal bar conditional formatting on the new column with the value turned off. It’s the world’s most compact horizontal bar chart, but it instantly shows the relative scale of the different numbers.
The second methodology ties into the last question as well, but it’s more about how I talk to business stakeholders. And, it’s a drum I’ve been beating pretty hard for the last 1.5 years. It’s a tactic, really, for capturing and qualifying hypotheses. Before I start any analysis, I frame up the hypothesis that I’m digging into by completing two fill-in-the-blank statements: 1) “I believe [some idea],” and 2) “If I am right, we will [take some action].” The power of this framing is that it forces an actual hypothesis to be articulated (the first fill-in-the-blank), and, equally importantly, it qualifies that there is at least one possible result from the analysis that would lead to some sort of business action. If I can’t fill in both of these blanks with confidence, I go back to the business and probe for greater understanding. It’s a huge efficiency driver, as the analysis itself gets time-consuming in a hurry, so it’s better to spend another ounce or two of effort on getting upfront clarity than to burn a few pounds of wasted analysis.
Q. What sort of attribution model are you using to allocate marketing funds?
Since we have virtually no formal marketing budget at Web Analytics Demystified, I’m going to act like a consultant and answer for how most of my clients do this. Multi-channel, cross-session attribution is definitely a hot topic…but most of them rely almost exclusively on last click attribution at this point as their primary approach. And, I think that’s okay. Getting last click alone can be messy enough as it is, and, if digging a couple of layers into last click data turns up oddities, then those oddities are just amplified tremendously when trying to look across channels and across sessions. What I do with several of my clients is rely on last click (even cross-session — last click for a specific channel generally “holds” as the last click if subsequent sessions are direct traffic) as the primary attribution mechanism. Then, as we think through and map out the marketing ecosystem, we’ll dig in to specific scenarios and use cases to see if first- and intermediate- clicks, as well as off-site impressions, were meaningful contributors. If so, we’ll adjust how we do the attribution. But, more often than not — with my clients — the volume of cross-channel clickthroughs to the site is a pretty small percentage of the overall traffic, so adding a lot of complexity to cover that small set of visitors isn’t worth it.
Q. Sneak preview: Please tell us a take-away that you will provide during your talk at the eMetrics Summit.
The macro takeaway is that analysts need to broaden the scope of their responsibilities. Most analysts are good when they’re given a question to answer and need to dig into the data to answer the question. The problem is that the right types of questions are a tiny fraction of what analysts get hit with day in and day out. I’m a Big Believer in process, and the overall theme of my session is that analysts need to develop and own a process that starts way before “the question arrives.” And, furthermore, their process needs to extend well beyond, “delivering the analysis results.” I’m passionate about it, so, as much as Jim likes to get people sitting near the front, I’m prone to getting worked up on the subject as I present. I can’t promise that a little spittle won’t fly from the podium!