Select Page

Blabbering About Marketing Analytics

In late September 2017, I participated on a panel at the Marketing Metrics & Analytics Summit in Chicago. A few weeks ago, I realized that the team at Insights Exchange Network had sent me a transcript of the session shortly after the conference. Hopefully, I was somewhat articulate for the audience without being too controversial. I decided to put these remarks here on the blog. The topics are a little more general than the typical things that we rant about here. If anything, they justify me still being the cynic that most of you have known for all these years. This information also gives a little different insight into how I think about analytics at Middlegame, but I should be careful: You might get a completely different answer from some of the other guys. Here are the questions directed initially at me:

MODERATOR: There is so much talk about automation in analytics and how artificial intelligence will drive that. How does automation fit into the structure of how you are approaching business problems?  Have you been able to overcome the “black box” label that we all dread?

TODD: I rely a lot on the extreme flexibility that comes from using Bayesian approaches to identify and organize the relationships and modelling coefficients behind the scenes, but let users have control of the model design. Leverage 21st century horsepower to let the users interact with multiple model designs in near real time, and give them ownership of the most important part of the black box, which is the hypothesis in question to be tested.

MODERATOR: There is an ongoing discussion on merging or integrating response results from marketing mix modelling and multi-touch attribution. However, the results can be pretty dramatically different. I know that you have worked in both. Do you have some thoughts on why the results are so different and how we could do better to reconcile them?

TODD: Multi-touch attribution is generally regarded as over-exaggerating the impact of marketing. From a modelling perspective, I think that a lot of these approaches haven’t included the concept of a baseline … shopper response without the marketing efforts (the touchpoints) being evaluated. There is a lot of opportunity to talk about data validity given all the questions that are going on with digital, but I’ll reserve that for data experts. My data issue is that marketing mix modelling has the baseline, since we are usually looking at a full market of shoppers and potential shoppers. Since we have put so much effort into targeting online customers, that definition of shoppers and potential shoppers is very skewed. I think that is huge, and really feels a lot like the pressure that built up in the ‘90s when evaluating direct mail. 

MODERATOR: The speakers and most of the conversations for the next two days will focus on what are the next big things to focus on for marketing analytics, but you seem to say that maybe we should also be thinking about the things that we seemed to skip over thanks to “big data,” “artificial intelligence,” etc. What do you think that skipped over?

TODD: We have put our current bets in analytics development in two places. The first one is emerging markets. I know that the economies of scale are quickly refuted there, so we are focused on how to use technology to use the existing data in these normally considered “data poor” environments to work. The second is trying to flip the “modelling currency” (business outcomes that we are trying to help marketers better address) back to all the things that are important besides sales. Obviously, the goal is to sell more. However, there is a lot going on in that statement. If we start to think about customers as the currency, we get to start talking about new customers versus retained customers, but also up-selling and cross-selling as the actual outcomes that add up to sales. The question is which combinations of media and message are the real drivers of these elements of sales. This is how we have talked in marketing for decades. Addressing these things now seems like a good idea given the computing power that is at our fingertips.

MODERATOR: We all know that the “big data” phenomenon has done wonders to bring analytics to the forefront of most business discussions. However, practitioners like us know that it also came at a cost to how we approach problems and try to bring better insights into how to solve those problems. What are the main drawbacks that you regularly run into when applying response modelling to these kinds of data streams?

TODD: I think that we have somewhat lost our way. My blowing off the fundamentals of marketing research–that was the precursor to analytics for me. Sadly, I think we have simply accepted that lots of data means analytics can have at it and have fun. The law of big numbers is a little different than that and we have honestly had to think about that since the beginning of my career when all the scanner data became available from retail point of sales data collection. The problem is that we are treating what is really unstructured data as structured data. The fundamentals of marketing research put experimental design and survey sampling techniques front and center. Basically, we know what the data is that we are applying advanced analytics to. Not so much with “big data.”  

MODERATOR: It seems like it has taken ages for analytics to be accepted at senior levels in many organizations, but we are really getting close. However, there is still a lot of resistance when companies can just hire a smart person who has a track record of brilliance, proven experience and obvious determination for solving things. Basically, the smart one doesn’t need analytics and decides how to dedicate resources to a plan already in their head. How do we get close enough to show that analytics can really help lead to a better answer?

TODD:  I find that most of the battle is not resistance to analytics because the topic is challenging for the economic buyer to grasp in the short amount of time they can dedicate to you. I think there has been a real awakening on that front. The uphill battle is often the all too quick response of “we tried that with another vendor and analytics just don’t work.” The problem is that the last supplier clearly over-promised on a solution that was half-baked at best and clearly underwhelmed the client in delivery. I think there is too much emphasis on immediately going to a turn-key solution and sadly that means meeting the lowest common-denominator out of the gate. Unless we are talking about purely transactional analytics (simple algorithms solved very quickly and subsequently many times per second) this is probably going to be a clear disappointment. That “didn’t work” conversation goes in circles when the truth of the matter is that you truly have a fully baked solution and you know it can help address their issues because very few problems are entirely new. I often have three or four examples of tackling the same exact issue (or something so close that differences are minor) over the years with different names and different faces. Instead of making those connections, we are constantly trying to overcome the barriers driven by a community that starts from the viewpoint of their solution as opposed to the customer problem. Again, this is nothing new. Marketing is riddled with this problem.

Thank you to the team at the Insights Exchange Network for taking out all the awkward pauses or any of those “umm”or “that’s interesting”responses that were really an attempt to get some time while I thought of my answer. I do recommend that you check them out at https://www.insightxnetwork.com/. Insights Exchange Network puts on a huge array of conferences that are all very meaningful to our industry and they do an awesome job. I look forward to attending my next one.

Middlegame is the only ROMI consultancy of its kind that offers a holistic view of the implications of resource allocation and investment in the marketplace. Our approach to scenario-planning differs from other marketing analytics providers by addressing the anticipated outcome for every SKU (your portfolio and your competitors’) in every channel. Similar to the pieces in chess, each stakeholder can now evaluate the trade-offs of potential choices and collectively apply them to create win-win results.