Analytics may be data-driven, but in any team, strong leadership is critical for making sure that the project fulfills an organization’s needs. I discussed data-related roles for a successful analytics team in my last post. The next set of people bring vision and direction to the project.
1) Analytics Professional: An analytics professional can use analytics tools and technologies to build the advanced analytical solutions. The analytics professional may have a focus in one or more areas – statistical, text, machine learning, optimization, simulation, graph theory and forecasting – but it is unlikely that this person has deep enough knowledge to develop advanced analytic solutions across the board.
Today it seems that analytical jobs require deep knowledge in R and Hadoop. I’ve found the best analytics professionals can use COTS analytic tools to derive deployable, repeatable analytic solutions. While there are no universal certification processes for analytics professionals, it is process in the making. There are academic programs designed to create these professionals, such as North Carolina State University’s Master of Science in Analytics. In 2012, INFORMS announced an analytics certification. Without a professional standard, many organizations have difficulty distinguishing between a person that knows how to use a software tool or a programming language and a person that knows analytics.
2) Senior Analytics Professional: I’m a big fan of double checking my work! A senior analytics professional must have inferential logic and the ability to scrutinize the work of other analytics professionals in the organization, making sure that the analytics that are developed are sound. This person should advise on the modeling approach, make sure that there aren’t any “leakers” that have been included in the model, make sure that segment of the population that has been included is correct, and even make sure that the top variables in the model aren’t an artifact of a poorly designed data preparation process.
You many find all this double checking a pain, but it can save you time and money if it catches a serious error. For instance, a dear friend of mine that was new to modeling called me one day to share the results of the analytic model for a cruise line offer that she had developed. She was delighted at the model performance and wanted to show it off to me! I took a look, and noticed that the model performance was too good to be true. In social science, if it is too good to be true, it probably is! The top variable for her model was “$ spent with the company in the last year.” The variable we were trying to predict was “Number of cruises in the last year.” If the person went on a cruise in the last year, they would have spent money with the company! While the model was terrific, it was invalid.
3) Analytic Thought Leader: “Thought leadership” is such a buzzword that it’s often hard to determine what it really means. From an analytics perspective, a thought leader helps the organization design and map analytic problems. Seems simple right? Well, that level of knowledge comes from a strong background in different analytic routines and research methods. It’s essential that the thought leader has significant practical experience using a mix of analytic routines and also a research-methods background so that he can apply the most appropriate methodology to the problem.
Generally, advanced degrees in the social sciences, applied mathematics, statistics or economics are good indications that the individual will have sufficient design of experiments skills to be a thought leader.
Think about the problems that you are solving; most of the time these are “people problems” — they come down to social science research studies. Without a broad base of knowledge, experience and education, the analytics thought leader may apply the same algorithm to every problem that he approaches. Or, he may apply an overly complex solution because it is the one that he is most familiar with, instead of focusing on determining the right problem to solve.
Communication skills for the analytic thought leader are a must. This person needs to be able to explain to the analytics professionals what analysis needs to be done, but also be able to talk to the business owners, interpreting the results of the analytics in a business context.
4) The Champion: Every team needs someone with the skills of a champion, someone who has enough clout to enforce changes and the initiative to not only find but resolve problems. In order to accomplish great things for the team, this person must have the power necessary to enforce process change. Without a champion, the true value of your analytics will likely be only partially realized.
Willingness to change processes is essential for a true champion – the results can be huge. Back when modeling for telecommunications churn was hot, one of my colleagues identified that a lost or stolen phone was a great indicator of customer churn. The SME knew right away that this was because of the time lag between losing a phone and getting a replacement phone. If a person with a lost phone no longer needed to fulfill the requirements of his contract, it was easier to go to the store and sign up for a new contract with any telecommunications provider, than wait for a replacement phone with the current provider. Because of this finding, the champion implemented an easy process for replacing lost or stolen phones with overnight shipping and no red tape. Sure, he could have just used “lost or stolen” as an indicator of churn, but by changing processes, this company was able to beat the other providers to a resolution.
Now that your team is in hand, focus on solving the right analytical problem. I’ll address considerations for problem-solving next on the Myths and Realities of Successful Analytics series. Meantime, download a white paper about making the most of your analytical talent.