Should we forget attribution modeling and embrace econometric media mix models? Here's a look at which you should start focusing on.
I've never been a big fan of attribution models and have always preferred econometric models that do their best to generate a practical media mix model.
I've explained my reasons in different ways to clients, prospects, and show attendees, but I doubt I've communicated them as clearly as Avinash Kaushik did recently in his SES NY keynote. Kaushik pointed out the lunacy of some of the attribution models being used by search marketers who think of themselves as fairly advanced. Also, I've always been a fan of monitoring bounce rates of landing pages as closely as one monitors eventual conversion to leads or sales.
One of Kaushik's now-famous pearls of wisdom regarding bounce rates is fully self-explanatory and never grows old: "I came, I puked, I left." Clearly, for most of us looking at any analytics program, it's boggling how high bounce rates can be, even for our most relevant and best-performing pages. Getting the bounce rate below 50 percent is doable, but it takes a lot of landing page tuning, copy testing, and layout adjustment.
If you take one thing from Kaushik's crusade for better user experiences, it should be "watch your bounce rate." While not everyone is capable of designing media mix and marginal attribution models, everyone has the ability to start improving bounce rates now.
Moving on to the topic at hand, let's discuss attribution modeling and how it's done wrong by nearly everyone who is doing it at agencies and in-house. As an economist by training, I've always pushed my in-house teams, clients, and conference attendees (and yes, you, my readership) to think about all media based on its marginal impact on sales (or, if you prefer, branding). By thinking "on the margin" (or what Kaushik calls, calculating "marginal attribution"), you can begin to understand the impact that each media element has in making the sale. That's beyond simple attribution modeling because this approach attempts to understand what the elasticity of your marketing and advertising ecosystem is with respect to each media element and touch point, including search.
Everyone in the search marketing industry has always known that there's a strong interaction effect between media and search. When it comes to media mix models, the interaction effect between search and other media (even, to some extent, other search media) is higher than the interaction effect between any other form of media. Think about it.
This fact should come as no surprise because people search as the result of a stimulus. That stimulus could be advertising (paid media) or unpaid media (public relations, social media, etc.), not to mention things like store visits and offline social media, also known as old-fashioned recommendations and conversations.
So forget attribution modeling and embrace econometric media mix models.
While econometric models can get sophisticated, you may be able to start thinking in a way that lets you act as if you are building an econometric media mix model, and part of that relates to the previous discussion on marginal attribution. To best allocate a fixed media budget or to best maximize volume given a specific measured (or measurable) metric such as sales, ROAS, ROI (define), leads, CPA (define), or net profit, you need to understand the cause and effect of changes to your media mix. That's different from arbitrarily assigning "credit" to media touch points when the reality is you're likely missing a ton of potentially critical touch points from your data set.
For most search and online marketers, we're only allocated online media budgets, plus, we have data relating primarily to online marketing, advertising, conversion, or influence. The areas under our control will typically include search, display, e-mail, and social media. Similarly, we generally only have control over the allocation of resources among those forms of media. While it's instructive to know about the rest of the traditional media plan (and perhaps the direct marketing activity), coordinating an experimental design across media in order to build a fully robust econometric model may not be practical for most search marketers.
Instead, what we really need to understand is the marginal attribution of media we buy to the conversions, as well as the elasticity of the media we're buying. Just because we've arbitrarily assigned influence value to one media option that was one of the media touch points, doesn't make it easy to maximize revenue (or, for that matter, profit).
Let's use a simple example of the differences between an econometric (media mix) model view versus an attribution view of a campaign. Let's stay within search and display and, for the moment, ignore the fact that for most marketers, more than half of the response to advertising comes from offline media and marketing.
If your site is currently in the third position in Google for keyword number one (cruise vacation), sixth position for keyword number two (cruises), and those positions each contribute the same marginal ROI to your bottom line as a display ad running in AdX, your ability to scale each of those ads up is very different. Raising the bid by 15 percent on keyword number one may result in no positional change, particularly in a competitive scenario. Doing the same for keyword number two may result in a jump of two positions and a commensurate jump in traffic. However, a 15 percent increase in bid price for the display ad running in AdX, Right Media, and several other exchange-type platforms results in the ability to spend six times more. Plus, the display ad may stimulate search behavior better than the head-term search keyword does.
Clearly, market elasticities, marginal profit contribution, and an understanding of how other media impact (stimulate) search behavior all need to be used together.
Welcome to Media Mix Modeling 101.
From here, it gets very arithmetic and very exciting. The Holy Grail is the perfect media mix model which includes offline media, but for now we'll settle for an understanding of the best media mix within search, display, video, e-mail, and social.
Pro Rank Tracker is a cutting-edge ranking tracking tool for keeping you up-to-date with all the latest changes in the rankings of your websites and videos.
Kevin Lee, Didit cofounder and executive chairman, has been an acknowledged search engine marketing expert since 1995. His years of SEM expertise provide the foundation for Didit's proprietary Maestro search campaign technology. The company's unparalleled results, custom strategies, and client growth have earned it recognition not only among marketers but also as part of the 2007 Inc 500 (No. 137) as well as three-time Deloitte's Fast 500 placement. Kevin's latest book, "Search Engine Advertising" has been widely praised.
Industry leadership includes being a founding board member of SEMPO and its first elected chairman. "The Wall St. Journal," "BusinessWeek," "The New York Times," Bloomberg, CNET, "USA Today," "San Jose Mercury News," and other press quote Kevin regularly. Kevin lectures at leading industry conferences, plus New York, Columbia, Fordham, and Pace universities. Kevin earned his MBA from the Yale School of Management in 1992 and lives in Manhattan with his wife, a New York psychologist and children.
US Consumer Device Preference Report
Traditionally desktops have shown to convert better than mobile devices however, 2015 might be a tipping point for mobile conversions! Download this report to find why mobile users are more important then ever.
E-Commerce Customer Lifecycle
Have you ever wondered what factors influence online spending or why shoppers abandon their cart? This data-rich infogram offers actionable insight into creating a more seamless online shopping experience across the multiple devices consumers are using.
October 13, 2015
1pm ET/ 10am PT
November 12, 2015