However, Forrester, Mobile Marketing Association, Gartner all agree that there are solutions.

The best overall view of the problems and solutions, in my opinion, is Forrester’s 2016 Measurement & Optimization Wave. It discusses the problems and analyzes 67 measurement companies, rating the top 10 companies and what makes them different: (reprint at http://www.marketingevolution.com/wave).

Mobile Marketing Association has successfully done a series of next generation attribution with marketers including Allstate, Coca-Cola, MasterCard, Walmart and more, publishing them for the benefit of the industry (here is peer reviewed article in Journal of Brand Strategy about the methodology and finding: http://www.mmaglobal.com/files/documents/stuart.pdf).

Gartner refers to it as Total Marketing Measurement. Their Marketing Leadership Council has a “best practice” based on Regions Bank sharing how to connect brand to behavior at the person level: https://www.cebglobal.com/

If you want to hear from marketer’s directly, Citibank’s expert, Mike Eichorst recently retired and is sharing what he learned from his 35 year career in measurement. He has tried it all. Mix modeling, attribution, person-level, design of experiments, you name it, he probably has tried it. The webinar is next week. I have the opportunity to interview him, and in our preparation call this week, he is going to be very specific about what doesn’t work and what does work. There will be time for audience questions too.
https://attendee.gotowebinar.com/register/7867468973690060035

In conclusion, almost everyone agrees that there are problems with both marketing Mix Models and Attribution, and most agree that the new generation of measurement, Unified Measurement or Total Marketing Measurement depending on the analyst terminology, is solving these shortcomings.

Your criticisms are an accurate reflection of the OLD generation of measurement. But, the good news is we are already in the next generation.

Dear “Critical Fan” (I am intrigued!)

Much appreciated answer. Your thoughts open the debate I was actually hoping for. They also help me understand that I was not that clear after all… So, to your specific points:

– I never intended to advocate for Marketing Mix Modeling, which I believe is built on very shaky grounds: a clearly identified, easily measurable outcome. The only reason for its inclusion was a reference to a potential use for aggregate data when success has been so stubbornly tied to a very concrete milestone.

– Fully appreciating the power of advanced statistics when it comes to “making up” for the lack of granular data (and fully embracing them at different levels) throughout the said “journey”, my contention is actually with their irrelevance in the face of isolation from the said defining milestone. In other words: we can use statistics to complete the picture given a large collection of “journeys”, but they cannot help nailing down a destination, for this one has a life of its own which is not measured through events or actions but instead disappears in the minds and private choices of our final target, consumers – we are missing an entire data set and I do not believe it is within reach.

– It is certainly a very good point that attribution is actually employed at operational level, as a means to rationalize specific investments. The problem being of course that not only does it end up influencing the strategic understanding of the marketing function, but it also assumes certainty in the attainment of the goal (often tied to the lifespan of a very fragile cookie). I do not believe it is a coincidence that most large corporations are these days facing a “civil war” between traditional “Mad Men” marketers (yes, resting in the laurels of a logic built for mass media reach, but truly familiar with the nature of brands) and performance marketing die-hards (aiming to turn marketing into a programming exercise, as that is their background).

– On the opportunity cost, you open a great debate. I agree, actionable improvements seem like the biggest priority. My question would however be: How can we be certain of a 10% improvement as a result of attribution when it is precisely the measurement of such improvement that we are calling into question? (rather than the methodology used to arrive at it through an enhanced combination of investments?). Which is why I am defending investments of any sort which accept multiple potential outcomes, with the optimization of sequential milestones (I have used “staged attribution” in a prior comment) appearing to be the low hanging fruit.

All in all, I believe that we have taken measurability to the extreme – as someone wrote recently, “Google Analytics has killed marketing”. What would be the point of the marketing function if we had a clear understanding of outcomes – could delegate it to a piece of software? Perhaps it is only paradoxical that I have to stand up for the unpredictability of human decisions/emotions after a considerably long career fighting old marketers in search for better answers 🙂

Thanks for your comments, Scott. I would actually agree that attribution tied to certain points in the journey can indeed bring clear benefits. We would in this case not speak of multi-touch attribution but rather “staged attribution”, as it obliges us to clearly define stage-specific goals, giving us far greater control over the timeline.

And yes, this is what I meant by “journey distribution”. I believe we are aligned.

Thanks, Mike. Very good point. We do suffer that scenario often in the B2B space. We happened to discuss it recently (during a Disruptive Digital Marketing NYC Meetup) and it was good to hear about a different approach to CRM leads which is more focused on “accounts” (with their many stakeholders downloading papers and researching info as a team) than “contacts”. We are in this case speaking of an “Account Journey” in which we must cluster all team interactions (cookies, requests, calls…) under a single path to (in this very case) a clear goal.

Very eloquent as always. Yet I feel the need to take issue with some of your contentions.

First and foremost, your case against advanced attribution is fundamentally flawed. By your logic, given that attribution (particularly digital attribution) doesn’t effectively measure everything that happens, we’re better off attributing using aggregate statistical tools such as Marketing Mix Modeling or viewing the world through the prism of a DMP.

1) You’re absolutely right that Digital Attribution platforms do not measure everything. However, they never intended to, and in fact, are modeled not to. Understanding the inner workings of advanced attribution models is sorely needed in order to understand their scope and outputs. While true that they attempt to view the digital journey at the most detailed level, various statistical methodologies are employed (such as hierarchical shrinkage, bias detection) in order to make up for the lack of observations. This means that Attribution models are simply machines with highly complex statistical models behind them, meaning that they too should benefit from the advantages of “Aggregate [not being] inherently inferior to granular data. And there is great power in correlations.”. This is important due to my next point:

2) We can agree that the role of attribution in the organization is NOT the end-game of determining ROI on marketing. However, the alternatives you posit (DMPs, Marketing Mix Modeling) does not take into consideration the operational realities of the enterprise. Attribution models fundamentally serve the needs of those that operate media campaigns and need to optimize marketing spend within their organization. For someone running a campaign, they need to know which creative has performed the best, how organic compares to paid, which keywords are performing and which aren’t. When $ millions are invested in online marketing, simply stating that MMM is a valid alternative to attribution modeling completely misses the actual needs of organizations to optimize their day-to-day marketing operations and get the most out of their spend. Believe it or not, attribution modeling is very well suited for this purpose. MMM is not. And here I circle back to my first point: Attribution is a statistical representation of marketing performance at the level that is operationally important to many marketing campaigns. And by your very own argument, we should agree that it doesn’t have to observe every interaction in order to be useful and relevant.

There are many other points to take issue with here – but my last point is that you take an “opportunity cost” view : if you implement attribution, you forfeit other investments. This is naturally true to some degree, but again I believe it is an oversimplification. We need systems that serve the needs of many parts of the organization. If your online marketing is able to (albeit only via statistical indicators) improve its efficiency by 10%, that should be a huge priority for your organization. If an MMM model shows that you can increase sales by 3% by investing in other channels, this should also be a priority. If via a DMP investment can improve retention rates or optimize other parts of your business, this should also be a priority. The question then becomes: how to chose which investment to make? The answer should be fairly obvious: the area where your organization is most capacitated today in order to take action and pursue improvements. That will be a discussion for another day.

Keep up the great work Sergio. Despite my criticism, I thoroughly enjoy your writing…

A “critical” fan

Great article and I admire you taking this stand. However, I would respectfully argue that “good enough” sales attribution and ROI mapped to certain points in the customer journey can bring real value to marketers.

I’m wondering if that is what you meant by “journey distribution” above?

For example, if you can track the click before a visitor becomes a lead, by mapping the click to the CRM activity, and then pull in the cost of that click, that would seem to have value. You then have legitimate tracking of successful lead generation. Furthermore, now that you have that customer journey point (we call it first optin), future sales can be tracked and attributed to the lead generation, giving ROI, or lack thereof, to your lead gen marketing efforts.

What is very interesting (to me any way) is how, in B2B instances, our channels for contacting businesses has changed over the years. I feel that the attribution fields in CRMs are saying more and more “website” – but may or may not have anything to do with successful SEO campaigns… it could be an in person referral that is now searching for your company or someone heard your radio spot and wants to learn more. First, last, middle – it doesn’t matter, because we will never know the mindset of the prospect.

One could suggest you use CRM data of leads, opportunities, wins as a guide to where you should invest, but that quickly becomes flawed. In most B2B sales you quickly move from a single point of contact – someone who researches the company, downloads the whitepapers, speaks with the sales people – to a boardroom of 15+ decisions makers from different backgrounds/specialties who ultimately have say on the purchase and who may or may not have ever interacted with one of your marketing pieces in a trackable manner, because the “researcher” in their company brought them all of the compiled information.

Just my thoughts.

Again well written article – gives me lots to think about.

Thanks for your input, Rakesh.

Mark,

Superbly articulated. I will be quoting you, no doubt! Many thanks.