Why Data Scientists Must Act as Mediators and Translators Between IT and Marketing

Cats and dogs have some serious communication issues. For one, cats raise their tails when they are angry, while dogs wag their tails when they are happy. A playful dog can inadvertently start a physical conflict for just waving his tail in front of a cranky cat, who’d say, “Do you know whom you’re raising your stupid tail against, dog?” Encounters like that cannot end well. I often see that IT and marketing departments are like that. They just don’t seem to understand one another. Techies often think that marketers are just too vague and touch-feely. Marketers think that technical people are too rigid, demanding logical statements during a casual discussion, like Mr. Spock in Star Trek. Let’s say that a marketer wants to target “high value” customers and other similar cohorts. To perform modeling and segmentation for such targeting, she would need representative samples of such “high value” customers. Now, a benign request like that can lead to a heated argument. I have witnessed many conflicts like that in person. From the technical point of view, a vague expression like “high value customer” means less than nothing. How are we going to determine that? By frequency of their visits? Online or offline, or both? Or, by total dollar amount? Total spending for life-time, or for a specific time period? Do we count Christmas seasons, or not? How about using accumulated loyalty points on top of all this? What about one-time big spenders who just went dormant? Are we still going to call them “high value customers”? While at it, how would one define the “dormant” period: 3 months, 6 months, a year or longer? This line of questioning may give you a headache, I know. But people who work closely with computers should think like this, as machine do not understand vague expressions (yet). Non-technical people may think that this whole process is tedious and painful, but defining specific parameters at the beginning of a project will prevent many mishaps and detours later. When conversations of this nature become heated, both sides would benefit from a mediator and a translator. That thankless job often falls onto an analyst who got caught in the middle. And that’s not a bad thing at all, as most analysts are equipped and positioned to be ideal moderators between marketing and technology worlds. What it also means is that good analysts must be fluent in both languages. Their main functions aren’t just technical and mathematical ones, but also a role as an interpreter of goals, strategies, methodologies and computing languages. For that reason alone, analytics teams should never be under IT or marketing departments. Unfortunately, companies tend to lump up teams that are seemingly technical into one large group under IT. Conversely, organizations that have been successfully monetizing their data assets maintain their data and analytics operation as an independent division under Chief Data Officer or Chief Analytics Officer. IT department would be in charge of system infrastructure and security, while data and analytics team would be responsible for data management and advanced analytics. Those two departments should be on an equal ground with distinctively different roles and responsibilities. One of the most common reasons that data assets do not fully materialize into tangible benefits is because information flow is disrupted among various constituents within the organization. One of the most common reasons that data assets do not fully materialize into tangible benefits is because information flow is disrupted among various constituents within the organization. Further, data is not something that stays in one place or in one shape, but fluid entity that must constantly be transformed and molded for users. Data would never gain any value on its own, and information business is not just about putting technical puzzles together. Armed with deep technical knowledge in data management and advanced analytics, heads of analytics are business-minded leaders who think through what project would be beneficial for the whole organization, how non-technical departments would employ analytics and how to increase overall decision-making efficiency. If the analytics team ends up reporting to IT or marketing department, a mediator who would objectively determine project specifications and priorities may disappear as well. The most difficult job in data management and analytics isn’t overcoming technical difficulties, but coming up with the right sets of assignments through methodical problem statement development. Novice data scientists struggle in the beginning of their careers, as in schools, professors would provide clear set of problems to solve along with pristine sets of test data. In real life, there is no one who is kind enough to give them homework like that, and worse yet, there is no such thing as a perfect set of data. Good data scientists, with a little help from their analytics leads, must be able to develop problem statements, define project goals, outline viable project plans, procure necessary data, refine obtained data, perform programmatic and mathematical functions, derive meanings out of data, summarize the results for end-users, recommend actionable next steps for decision makers, systematically share knowledge, and continuously improve circulation of information. I’d say that is some job description, and it’s not always technical. The analytics team must remain independent, as an analyst must work as a translator and a mediator between IT and marketing teams, who are often like cats and dogs. It is a lonely and tiresome position in between vastly different ways of thinking, but to be successful, data players must accept that role wholeheartedly. Too bad that many executives do not recognize the intricacies of this balance, and boldly lump what they don't understand into one big technology group. But analytics teams rarely thrive under IT or marketing department. If you doubt that, have a candid conversation with one of those data nerds caught in between the two worlds. Inside Machine Learning and Predictive Analytics