But here’s the rub: MMO has been largely opaque to its users, and backward-looking in its orientation. To make MMO truly effective, CMOs should demand an approach that is transparent to their organizations, forward-looking through test-and-learn sequences and focused entirely on results. Done thoughtfully, MMO can help marketing organizations realize cost efficiencies of 10% to 25%.
MMO has already been through two distinct phases. “MMO 1.0” built models from the top down with aggregated data to explain how spending drove outcomes. Unfortunately, limited variation in spending often made these models statistically weak and, as a result, insufficiently persuasive to motivate change. “MMO 2.0” works from the bottom up, adding fine-grained data on individual customers’ behaviors to more accurately assess the impact of various media, such as the effect of display ads on paid-search return on investment. While MMO 2.0 has yielded benefits, its complexity prompted most CMOs to turn to outside firms for help, and many find themselves dependent on blackbox models that hide their data and logic.
Now we are seeing the advent of “MMO 3.0.” This phase promises better results through artificial intelligence (AI)–powered decision making, based on improved software and newer and even bigger data sets, such as unstructured text and digital exhaust from the Internet of Things.
What should MMO 3.0 look like? With apologies to the movie Fight Club, “The first rule of advanced analytics is, we don’t talk about advanced analytics—we talk about results.” To get results, Bain & Company’s research and experience suggest that marketers need to define and realize this next phase of MMO differently: with less focus on algorithmic sophistication, per se, and more emphasis on five surrounding conditions to create a dynamic MMO (see Figure 1).
Condition 1: Strategic alignment
In any good analytic effort, the first task is to ask the right questions. Smart people will argue passionately about where to focus (what goals, which parts of the business), based on their own facts, logic and biases. If you don’t have a way to get them on the same page, expensive analytic resources will be misdirected at best and paralyzed at worst. We have found it useful to ask:
- What is the customer’s journey, from awareness through loyalty?
- What channels are relevant to this process?
- How is our marketing mapped onto this process?
- How does our marketing perform at each point (traffic, efficiency, conversion)?
- Based on trends and benchmarking of these metrics, where are the bottlenecks in our performance?
- What could we do differently, and what would the impact be?
Condition 2: Access to data
Here’s a cautionary tale. A major bank hired a firm to build an MMO model. About 80% of the project budget went to gather the necessary data and make it usable. The model might have been sound, and the insights useful. But since it was a one-off project, the vendor couldn’t spend time to document or explain the data, or how it was gathered and transformed, or how to get to it the next time. When the vendor presented the model’s answers, even the few executives who understood the results didn’t believe them, because the data inputs had been explored in the vendor’s lab rather than through a well-communicated process. The sad result: The bank found itself locked into the arrangement because only the vendor knew what was going on inside the box. Uncomfortable with that arrangement, the bank discontinued the effort and so returned to the data dark ages, waiting for another day to take a new run at MMO.
Two actions can mitigate such problems. One is to inventory, document and secure proper access to data sources independent of any particular modeling effort, and to ask any vendor to fully support this. The hardest, most underestimated part of an analytic effort is the data engineering needed. It’s much less expensive to keep this work going once done than to rediscover and redo it episodically. The second useful action is to make the exploratory data-analysis phase a transparent collaboration between analysts and marketers (which complicates the case for outsourcing). This approach enables everyone to understand the models that emerge in subsequent phases of the analytic process. Even when those models move into more advanced machine learning or AI realms, where “feature selection” (choosing significant variables) may be more opaque, watching for changes in the broad patterns of the data that feeds them helps to prevent nasty surprises.
Condition 3: A balanced approach to analytics
There’s a joke about why the man who lost his keys in the bushes was looking for them under a nearby streetlamp: Because that’s where the light is! One risk with analytics is being constrained by the data and skills you have, rather than what you need to answer a question properly. As mentioned earlier, existing data may lack sufficient variation, or be irrelevant to current market realities. Analytics, research and, in particular, testing are all present and well balanced in the best MMO efforts. We recommend embedding test plans in campaigns and programs by default, rather than treating them as episodic add-ons.
Condition 4: Operational flexibility
MMO leaders balance investments in generating insights with putting them to work. This entails a marketing operating model and technology infrastructure capable of acting on insights, at the level of grain and efficiency the insights demand. For instance, if we discover through web analytics that customers who take action X convert better, we need to be able to retarget people who did X.
Condition 5: Analytic marketers
Much of the debate about the human dimensions of MMO lies in how to develop an ongoing process, not just an episodic project, by addressing questions such as how much to centralize. Our experience suggests that successful organizations tackle this issue last, not first. The answer remains fluid and tailored to circumstances. MMO leaders start instead by examining and working on the interactions among analysts, marketers, IT, agency partners and other vendors, all disciplined through regular reviews of performance. A practical formula emerges to guide this engagement and collaboration: “Shared facts + shared logic + shared interests = Better outcomes.” Given that most organizations are still in early days with MMO, they would benefit from a “center of excellence” structure.
Improving through a portfolio approach
Many organizations approach MMO as a capability-building effort. They treat it as a project, with tools and processes to implement and then run. By contrast, we have found that practice makes perfect: MMO leaders work to improve the conditions we’ve described indirectly, in the context of specific marketing campaigns and programs oriented on business results, rather than directly, as capability-development initiatives (see Figure 2). Further, they start simply, through transparent, sustainable iterations that go through complete cycles, including data-driven hypotheses, crisply executed action and then careful consideration of feedback. They don’t just shoot first and ask questions later.
Analytic teams often conceive of their MMO agendas as task lists of passively received requests for studies or reports that at best are FIFO (first in, first out) but more often SWGG (squeaky wheel gets the grease). A practical alternative approaches the MMO agenda as a venture investment portfolio. The portfolio combines regular performance expectations (perhaps quarterly, though this may vary for different organizations)—notionally, three insights, two tests, one scaled solution—with quarterly governance and priority setting from the senior executive team.
At the investment management firm T. Rowe Price, Paul Musante, head of client and market insights for US investment services, developed a learning agenda with his senior colleagues based on a regular, structured review of market and business performance. That, in turn, serves to discipline and filter the inevitably large demands placed on his teams. “It’s crucial that we allocate these scarce resources carefully, not just for insight, but for where we can practically act as well,” he said.
To maintain confidence that you’re on the right path toward results, recall the fable fragment from the Greek poet Archilochus: “A fox knows many things; the hedgehog, one big thing.” Be a fox: Think about customers and shifting bottlenecks in their experience. Stay practical and in control of your data. Use balanced analytic approaches. Don’t let analysis get too far beyond action. Cultivate analytic marketers. And focus on incrementally better insights and predictions that you understand, rather than big-bang black boxes you don’t.
Cesar Brea, Laura Beaudin and Andreas Dullweber are partners, and Brian Dennehy is an expert vice president, in Bain & Company’s Customer Strategy & Marketing practice. They are based, respectively, in Boston, San Francisco, Munich and San Francisco.