Skip to content

Analytics Professionals – Must They Be Good Communicators?

Most job ads in the technical arena list communication among the sought-after skills; it consistently outranks many programming and analytical skills. Is it for real, or is it just thrown in there by the HR Department on general principle?

The founder of a leading analytics services firm told me that good biostatisticians in the pharmaceutical industry don’t need to be good communicators. No matter how poor their command of English, and no matter how high their geek factor, they will be listened to and, though repetition may be needed, they will ultimately be understood. Why?

A CEO running clinical trials that cost up to a billion dollars, and that represent potential revenue many times that, will make sure that the biostatisticians designing the study, performing the analysis and submitting the results to the FDA are attended to. The latter need not clamour to make themselves heard, or sell the CEO on the value of statistics. The mountain will come to the statistician, since ignoring the statistician invites calamity.

But this is the exception that proves the rule.

Most analytics and data science projects fail

Let’s start with the fact that a significant proportion of analytics projects in business fail. One executive with Alteryx, a vendor of software to manage the data science pipeline flow, estimated that over 80% of analytics projects are implemented only partially, or not at all. Failure of statistical or machine learning models is not usually the issue. Speaking with leading consultants and team leaders over the years, I find that a consistent theme is that the most common reason for the failure of analytics projects is not technical failure, but rather system failure. Different parts of the organization don’t talk to one another and commit to a common strategy.

One leading consultant described for me a huge project in which leadership decided, by edict, that rules-based vetting of transactions for compliance with U.S. and international standards would be replaced by model-based approaches. In other words, transactions would be turned down or flagged not by virtue of triggering one of several rule thresholds (the established system), but on the basis of a score from a predictive model. The consultant was, of course, pleased to hear that statistical models would be developed, but had to caution the client that the shift in approaches had to be one in which multiple stakeholders (including regulators) had to be involved. The strategic vision would need to be articulated and developed in a collaborative fashion throughout the organization. In other words, the real challenge was institutional, not statistical.

Dean Abbott, a noted analytics consultant, compares successful implementation of analytics to a 3-legged stool:

  • Business and domain experts: if they are missing, the analytics team may solve the wrong problem;
  • Analytics team: if they are missing from the problem-defining phase, the problem may be defined in a way that is not amenable to an analytical solution;
  • IT/database: if they are missing from the problem defining phase and the implementation, the project may not get the data it needs, and it may falter at the deployment phase.

Field Cady, author of the Data Science Handbook (Wiley), adds another group:

  • Software engineers: if they are missing, the solution that is developed may not be deployable in the way that the business situation needs

So, yes, communication is important! But not everywhere in the same way.

Companies recently built on a foundation of data and data science (Google, Facebook, etc.) will have a shared basis for collaboration and communication. Cross-functional communication is still required, and the analytics team must be attuned to the business context of their project. But the people on the business side are more likely to be part of a tech culture that provides a common basis for communication.

But for most non-data, non-pharma companies – i.e., the bulk of the economy – there will likely be a communication gap between the data analytics people (who may be newer) and the goods and services people. And analytics and data science, while important, may serve mainly to boost profit or reduce costs – data is not the fundamental core identity of the business. For these companies, the communications burden lies with the analytics people.

Field Cady has these communication tips for analytics people who want their projects to be a success in a wider organizational context:

  • Know your audience. Put yourself in their shoes – a business person will be lost unless the business goal of the project (not the technical goal), and its potential impact in dollars, is evident.
  • Make it concrete, but not technical. Abstract concepts and general principles may be useful to frame a discussion, but you must translate into specifics and examples for the discussion to progress and be useful. Moreover, sticking to abstractions is often just a lazy attempt at evasion, and can indicate that you don’t really understand the problem.
  • Don’t flaunt the math. Math is not synonymous with clear thinking, says Cady. Most predictive modeling, and much of data science does not require high-end math. In almost all data science projects, the beginning (specification of the business opportunity and the resulting technical problem) and the end (architecture of the solution) must be in quantitative, but not necessarily mathematical, terms. Mathematics comes into play in the middle, as needed, to accomplish tasks faster or more efficiently, but the specific goals of the tasks can be understood without the math.