4. People and Organization—Cultivate “Analytic Marketers”

The dominant tendency in marketing organizations is to draw charts first and then define and hire narrow roles into them. This is because the range of functions to be encompassed in a marketing organization is broad, and the knowledge required to execute each can be somewhat specialized. Inevitably, this leads to a situation where marketing analysts are uniquely defined and functionally segregated from other marketing roles. What follows is a too-common pattern: More numerate marketing analysts generate reports that less numerate marketers often don’t use. In short, Venusians and Martians remain on their respective planets, without communicating effectively.

The advent of digital channels and the data they offer has made numbers somewhat less escapable. In “pure play” digital firms, numbers are oxygen for all marketers. Consider the case of the online home furnishings retailer Wayfair, where, in 2013, CEO Niraj Shah said that half his marketing staff knew how to use SQL to query the firm’s data for the information they needed. In firms where digital channels have been developed on top of pre-existing channels, progress is somewhat slower.

In Pragmalytics, I suggested that the “org-first” approach to designing marketing organizations was precisely backward. I advocated for attracting strong, numerate generalists, and for providing training and constant application of numbers-driven marketing for both newcomers and existing team members. One way is to insist on having conversations across marketing and sales functions at the lowest possible plain-language common denominator. Both of these functions have, for many reasons, and not uniquely, evolved jargon-rich languages to which even experienced practitioners struggle to adapt when they move from shop to shop, since the same thing might go by several different situation-specific labels. For example, what, specifically, do we include when we say “SPIFF” (where does a short-term sales incentive stop, and become part of the broader comp plan?) or “DSP” (precisely what functions are we supporting through our digital advertising demand side platform)?

1. “Statistical Analyst Salary,” Indeed.com, http://www.indeed.com/salary/Statistical-Analyst.html.

However, a more illuminated understanding of the analytic process would drive toward recruiting and cultivating a different set of people. First, it’s important to recognize that only a small part of a marketing and sales challenge involves developing descriptive models and predictive algorithms. A lot of time is often consumed beforehand in gathering and staging data, examining trends and distributions visually, and packaging any relevant information and insights for communication and application afterward. Second, as we discussed in Chapter 3, “Operational Flexibility—Don’t Analyze What You Can’t Act On,” a smart analytic strategy starts simply and doesn’t expand any faster than a firm’s ability to absorb and execute on insights it generates. Accordingly, your staffing efforts should recognize and follow this pattern, focusing more on smart, analytic generalists who can learn a range of directions and then leverage specialized skills at the edges of the work, via small internal cadres or external partners, rather than trying to stuff these resources in at the core and have analytic Ferraris running in first and second gear all the time and burning out their clutches.

Virtually all the conversations for this book bear this suggestion out. Belinda Lang, former CMO of Aetna’s consumer business and a long-time executive at American Express, who at one point managed a team of fifty statisticians, valued “entrepreneurial” as much “analytical” in the people she recruited for her teams, who, coincidentally, included the future Harrah’s CMO (and CMO Magazine 2010 CMO of the Year) David Norton. David himself noted how the ability to understand context and implications around analysis is a very important success factor there. He describes the head of his statistical modeling group as “a brilliant guy from the University of Nevada at Reno, who really was as much a strategic analytic marketer as he was a statistician. He really understood the marketing realities behind the numbers.” Ben Clark at Wayfair notes significant levels of cross-disciplinary fluency: statisticians who can program, programmers who get models.

Nonetheless, reality intrudes. In a world where the renaissance generalist comes at a premium, if you can find one, Melanie Murphy focuses on finding well-qualified specialists—say a modeler and a database specialist—and teaming them closely and carefully. But even she screens for the ability to get your own data and communicate what your analysis means. The moral: Don’t relax your recruiting approach or standards, even as you are creative about how your overall team is put together.

2. Wayfair has its own internally developed curriculum; you might start here, with the very accessible one provided by W3Schools: http://www.w3schools.com/sql/sql_intro.asp.

Assuming you can wrangle the data, next up is the ability to analyze it. In Chapter 1, “Strategic Alignment—First You Need to Agree On What to Ask,” we describe a basic analytic pathway any executive can apply. Beyond these skills lie techniques we rely on the pros to provide. However, as we look for and develop these, Scott McDonald’s experience at Condé Nast suggests it’s important to stay aware of potential biases that come with academically trained statisticians you might think to hire. As Scott puts it, “I came up as a statistically-trained sociologist, baptized in the Church Of The Random Sample and educated in The School Of The Normal Distribution.” But, he observed in our conversations that the base case taught in academic statistics settings—random (unbiased) samples producing bell-curve distributions—is less common in the world of sales and marketing. The problem, Scott observes, is that many analysts today that grapple with big data sets start, rather blindly, with summary statistics about data sets that implicitly assume the academic base case, rather than the marketing and sales realities of biased samples and skewed distributions.

Scott pointed to Nate Silver’s analysis of recent elections as a good example of the “arbitrage” these assumptions offer. By studying and handicapping the sampling biases of various polls, Silver was able to call each election right down to the very electoral vote count. In our conversation, he suggested how his analytic approach identifies and adjusts for these hidden biases. His ideas offer the basis for useful screening questions (“Tell me about the data you worked with—what assumptions did you make about it, and what adjustments did you have to make to it, in order to understand what had happened and make predictions from that?”), or a training curriculum (“Here are the characteristics of data sets we encounter in our corner of the sales and marketing domain...”).

Marketing and sales analytics should of course be construed more broadly, extending to research and testing as well. There’s a tendency, as we discuss in Part II, “Practical Analytics: Proven Techniques and Heuristics,” to organize these capabilities in silos. If each group is then left to develop on its own, there’s a risk that it might recruit past the point of diminishing returns in its silo. At any given point in time, the best marginal hire might be in another analytic discipline. For example, your next best move might not be a Ph.D. statistician, but rather a recent MBA with two or three years’ experience designing and executing tests. In Part III, “Making Progress,” we describe an approach to managing a portfolio of analytic initiatives that can help you decide who your next hire should be.

Here’s a short checklist of skills, traits, and experiences to look for across your team, if not in an individual analyst:

• Database skills—Can he or she source, clean, model, and stage data to subsequently analyze?

• Analysis skills—Can he or she describe data through trend and distribution analysis? How about multivariate and other sophisticated statistical analysis?

• Subject matter expertise—How well does the analyst understand the dynamics and vagaries of the business domain the data describes?

• Communication skills—Can the analyst find and support stories out of the data and analysis? Are these stories tied practically to available options, and do they advance decisions or slow them down?

• Management skills—Can the analyst get the job done, on time, on budget? Can he or she lead a team in doing this?

• Practicality—Does the analyst demonstrate focus on the decision or action to be informed, and can he or she distinguish “good enough to answer the question” from “perfect”?

• Curiosity—Does the analyst generate and test hypotheses beyond what’s given?

But beyond hands-on practice, where do analytic leaders turn for ideas? Bed Bath & Beyond’s Melanie Murphy says, “I don’t read analytics books; I try to read books focused on leadership!” One title she’s found useful is Tom Davenport’s Competing On Analytics, especially for the way it positions the capability she manages as a differentiating asset that senior executives should cultivate, rather than the cost center it’s traditionally perceived to be. Talbots’ Rob Schmults says, “If I want to learn about something new, I’ll root around to find someone who’s trying it too, and call them up! The verbal exchange is usually the most efficient for getting to the heart of it. Plus you can get lessons learned and pitfalls to watch out for that you won’t read anywhere or hear on a stage.” That’s an unusually extroverted instinct for many analysts, so any development program you construct might usefully weave in attending a few conferences, and refuse to reimburse dinner for anyone who doesn’t come home with at least two business cards, along with thorough trip reports of what’s been learned. Or, you can reward the most cards, best write-up, and so on, depending on your motivational inclinations.

• How much do we centralize this function and have it serve multiple business units and functions, versus pushing it into each group?

• What belongs in “the business” versus in IT?

• When and how should we combine “marketing and sales analytics” with “business reporting”?

• Speaking of reporting, to what degree should supplying ongoing reports be combined with analytics groups?

• What degree of integration should I have among my research, analytics, and testing groups?

• What should I try to do in-house, and what should I ask external partners to do for us?

Of course, there are many ways different organizations have answered these questions, and the answers they come to are influenced by a number of factors, including:

• The analytic needs of the business (finding new ways of competing in a choppy and strategically uncertain context, versus riding a successful horse as long and as efficiently as possible)

• Where in a particular technology wave the firm finds itself, and the attendant supply / demand dynamics of that particular phase

• Where a firm is physically located, as this affects its ability to find and attract people with different skills

• The individual skills of its people, as sometimes it makes sense to centralize and share a small but strong analyst cadre, while in other situations it makes sense to push analytic support to talented managers operating at the front edges of the business

• The relative leverage from different analytic approaches at various points

• Factors relating to how effectively insights can be communicated and absorbed across the firm, including distance, language, and culture

Even controlling for these differences in circumstances, the answers appear to be less black and white, and more a matter of degree. The problem is that most organizations don’t do nuance very well.

There is a different way to answer questions of organizational structure. It starts with facing and taking advantage of two realities. First, each business opportunity may require different analytic skills, or at least different blends of ones you have. So, the optimal organizational structure may be situation-specific. Second, the circumstances discussed above change, and with them changes your organizational answer. Your killer drug goes off patent. You expand into a new country. It finally really is the Year Of Mobile. Doug Cutting and Mike Cafarella write Hadoop, and Nate Silver joins your summer intern program.

Of course you try to keep an eye on all these changing circumstances. But looking at “input drivers” alone would crush you with its scope and complexity. A second way, then, to manage organization is to track outputs as well as inputs, by looking at the rate at which valuable insights are identified, proven, and scaled by the whole organization. In our work with clients, we begin by grouping demand generation challenges and opportunities into a portfolio, which we mine in quarterly cycles for “3-2-1” results—notionally, in any quarter, our work needs to yield three fresh “news you can use” insights, two real-world tests of these or prior insights, and one idea that’s promoted to “production.” (The quantities are arbitrary; they’re not a bad place to start, but your circumstances may dictate that they be lower initially and higher once you gather steam.)

There are three crucially different characteristics of this approach. One is the integration of the idea that insights need to be not just generated, but tested and scaled as well. Putting these metrics alongside pure insight generation at the heart of the program helps connect analysts to operators. Second is the relatively short timetable (a quarter) for inquiry. As we discuss in Chapter 13, “Culturelytics—A Practical Formula for Change,” cultures are collections of values, and values are generalized from observations of successes, failures, and their causes. If you want to improve results and believe that changing culture is part of that, you need to accelerate the rate of observations off which values and cultures change. Also, by doing that, one side benefit is that you take the pressure off any particular change you make to be The One. The third distinguishing feature is its focus on continuous improvement, even as the rate of that improvement may fluctuate. Yes, you may have benchmarked and established goals in the context of customer needs, competitive performance, and necessary returns for the business, but whether you’re near or far from those goals is less important than progress toward them. Imperfect, temporary organizational changes that make you better are more valuable than stalling outputs while you perfect structure. (We explore these issues and approaches further in Part III.)

Our executives’ experiences and preferences for organizational structures demonstrate this diversity. Doug Collier, CMO at La-Z-Boy, has IT reporting to him, and outsources (for now) marketing mix models and attribution analysis to MarketShare Partners. Rob Schmults at Talbots prefers to insource analytics; in his experience, getting it from media partners produces narrow answers or, worst case, biased ones. Judah Phillips and Melanie Murphy believe strongly in centralizing the analytics function. Paul Magill at Abbott and David Norton at Harrah’s created highly matrixed models with significant accountability to and direction from “the field.” Mo Chaara’s Corporate Analytics Group at Lenovo represents an effort to drive some critical mass for specialized expertise from the center, even as significant analytic capabilities remain aligned with and reside in different business units there. Belinda Lang’s experiences at American Express included running a “business interface group” in the IT organization that leveraged fifty statisticians to provide valuable services, but still needed to bridge cultural differences across organizations to be effective, despite Belinda’s considerable personal credibility and extensive relationships.

But look more closely. In Doug Collier’s case, organization is partly a function of his significant knowledge of and comfort with IT. His CEO is more comfortable having him own it, and IT is comfortable taking direction from him. Rob Schmults, who if not a father of ecommerce was in the room when it was born, is better equipped than the average client partner at a media agency or modeling firm to evaluate the results of an investigation. Judah Phillips and Melanie Murphy are both serious subject matter experts, whose skills and experiences serve not only as management for their analysts but as the fulcrums for strategic alignment and objective decision making in their firms; with strong capable leadership available from them, there’s significant leverage to a centralized model. Paul Magill and David Norton had to wrangle central leverage in highly decentralized firms where real power lay in the hands of the country manager or each casino’s boss. In every case, each of their solutions was a compromise between vision and reality that allowed each them to build momentum toward the visions and goals their firms had.

Some discussion questions:

• Are your people consistently able to answer the analytic questions that are put to them?

• If not, why?

• What skill deficiency comes up?

• What organizational barriers emerge?

• What’s your informal networking plan for shaping roles and generating leads related to skill deficiencies?

• Which social networks and conferences will you invest time and funds in?

• If you’re thinking about adjusting your organizational model to overcome a barrier, is your analytic leadership up to the challenge of managing through the change and beyond?