The Missing Link: Why Some Analytics Projects Struggle to Get Traction and How You Can Do Better

I was trained as a scientist. The idea that sound data and well-constructed studies will speak for themselves and the right actions will naturally follow was deeply ingrained in me. After 20-some years of practice, I realize that I have been wrong on that point.

Current practice in analytics and data science generally falls into three broad buckets of skills. First, you need content expertise – knowledge of the quirks and attributes of the domain under study. This can include insight into good variables to use for measurement and interrelationships among variables. This might be years of experience in marketing, or HR, or teaching kindergarten, or predicting election behavior or any other specialty. Without that content expertise, you can end up with results that are nonsense. Check out spuriouscorrelations.com for examples of very tight, yet utterly meaningless relationships. (My personal favorite is the .8 correlation between the letters in the winning word in the National Spelling Bee and deaths by venomous spiders.)

Second, analytics projects often need someone to wrangle the data – they need data expertise. How do you get your hands on the needed data, extract it from its current cozy home, clean and transform it, and tie it to other related variables? This is where all your SQL database skills and Python scripts for scraping data and NLP engines to sift through unstructured text come in. The efficiency these tools provide us is astounding. These skills are critical to the success of many analytics projects.

Third, analytics projects require analytics expertise. This includes depth of expertise with multiple analytical methodologies and sufficient knowledge to choose appropriate ones. It requires good thinking about samples and datasets, and the ability to convey results in an understandable and compelling way.  Analytics is in itself a vast and wondrous domain, with lots of subspecialties and vigorous rivalries between adherents to different analytical schools of thought. 

Together, these three broad skill domains form the core of most analytics work. Yet, even with all this robust skill, projects often go nowhere. While surprise reveals can make great television, I’ve found it to be ineffective at work, especially when the core message of the reveal is “I am so much better at your job than you are!” Teams get caught in endless cycles of “fetch me a rock.” Stakeholders ask “Could you add this one extra variable?” which generates 80 hours of rework. Worst of all: results are met with the dreaded response: “That’s interesting” and no action is taken on the data findings.

In the past few years, I’ve had an insight: the missing link in getting results with analytics projects is not in the quality of the analytics projects themselves – the missing link is influencing expertise. Now, as a scientist, this is a little tough for me to swallow. Good results stand for themselves! As I’ve grown, I’ve come to see influencing is different than convincing. Convincing is simply arguing for the rightness of my ideas. Influencing should not feel “Sales-y.” Instead, influencing starts at the very beginning. It begins with a partnership, jointly identifying important problems to solve and agreeing that a particular approach or family of approaches would be useful in guiding decision making. Influencing includes identifying the key insight from the analysis that will be useful to the project’s stakeholders, and the ability to frame that insight relative to the decisions that need to be made. The ability to identify actions to take based on those insights – actions that are feasible, will be effective, and that stakeholders will endorse – is an influencing skill. These actions are crucial to successful projects – projects that are relevant and realistic, have solid “pull” from stakeholders, and are laser-focused on yielding improvements for their organizations.

All four of these families of skills are essential for analytics success. Each of them is needed at multiple points throughout any given analytics project. Analytics professionals are likely to be deeply competent at one or two of these four skills – vanishingly few can do all four well, and that makes sense as they all require years of sophisticated practice.  Analytics professionals must, however, be appreciative of the contributions from colleagues who are experts in complementary areas. And they must be able to effectively partner to engage the full spectrum of required skills which will enable successful outcomes of their analytical work. It takes a village to do analytics well. Learning to add influencing skills to your portfolio of capabilities as a professional and as a team will help you deliver more compelling, more actionable projects.