The ONE Overlooked Skillset that Causes Most Analytic Projects to Fail

When I tell people I am an Analytic Translator, the most common response is: "Wow, that’s a really vital role. It’s so important to have someone who can explain complex results.”

That’s true; knowing how to interpret analytic results and deliver them in ways that audiences understand is critical.

Talented analysts work hard to select the right information and portray it with visuals and language that recipients can grasp easily.

It’s a specialized skill that, for good reason, gets a lot of attention. For example, the online education platform Coursera lists over 500 results for “Data Visualization.”

Amazon lists over 5000 books results on the topic.

Here are a couple of my favorites: one, two.

And, not surprisingly, visualization is included as an essential component of most data science curricula. 

Data scientists know this translation skill —from analytics to business— matters.

But that’s not the most important role of analytic translators.

There is a different skill overlooked by many data science professionals (actually, it’s not limited to this profession, but that’s our focus today).

The ability to communicate with business professionals in a way that elicits, clarifies, and defines their exact analytic needs.

From experience, I suspect readers will react to my description of this skill in one of three ways.

1. One group will presume it’s not relevant to them because they already “communicate all day, every day. It’s not hard.”
2. Another group will not be interested in this topic because human interaction is nowhere near as fun as analyzing data and they don’t intend to venture away from their computers anytime soon. Both are common, understandable reactions.
3. But a select group of readers will be intrigued and wonder whether there is a hidden communication advantage that can advance their careers.  I’m talking to you.

Why we should care

Anyone who works in business analytics knows that projects often fall short of their intended business objectives.

It happens.

We’re so used to it, maybe we don’t realize how often.

While there are limited studies of the exact rate of failure, we can find some indicators.

• Gartner, a $4B company that provides insights to corporate executives, projected that only 20% of big data analytic projects will deliver value to business through 2022.
• Business journals highlight the level of disappointment with analytics: Why so many data science projects fail to deliver (MIT Sloan),  Companies are failing in their efforts to become data driven (Harvard Business Review), and 10 Reasons Why Your Organization Still Isn’t Data-Driven (Forbes).
• In a 2021 survey of corporate executives, over 90% responded that the biggest barrier to successful implementation of Big Data was people, communication and culture, not technology.

The basic message: failure is common. Whether we believe the failure rate is 80%, 50%, or even 30%, it has huge implications.

What specifically goes wrong?

Having worked at the interface of data science and business teams for 35 years, I’ve witnessed thousands of interactions where the business expresses a need for information and the analytic team decides how to fill that need.

In very few cases do both parties take the time to get really clear about what needs to be accomplished, how the information will be used, or who the audiences will be.

Instead, each team assumes (or guesses) they know what the other means.

Then, analysts proceed quickly to gather data and apply methods to answer what turns out to be the wrong question.

Analytic translators know there is a critical step BEFORE creating a detailed analytic design:  listening and asking questions that help the businessperson get clear about their own thinking and get the big picture.

Here’s why.

Ideas rarely arrive fully formed

The first thing someone says is rarely what matters, and sometimes not even close. Considering the pace of business today and the rapid-fire, back-to-back meetings on multiple topics we cannot expect a complete idea the first time.

What they ask for first is not likely what they need. Not because they don’t know but because they haven’t had a chance to think it through.*

If we answer their first question, it’s not likely to meet their real need.

With a brief dialog and some guided prompts, thoughts crystalize.

It’s not unusual for the fully-formed idea to be quite different from initial stated request.

Context will impact approach

When I consult with analytic teams, it is not unusual for them to shrug when I inquire about the purpose of the inquiry. Understanding how the issue came up and how information might be used should influence the way (or possibly if) a project gets done.

Imagine the question is: what are the strongest predictors of vehicle accidents among our truck drivers?

Certainly, we can gather many sources of data and build sophisticated models to answer the question.

 But consider if the business wants to use the information to:

• Simply understand the factors better
• Make policies about scheduling and duration of routes
• Invest in intensive (expensive) driver training
• Hire lower-risk drivers
• Make a decision about what type of vehicles to purchase

Decisions about data selection, model specifications, and level of acceptable predictive inaccuracy should consider how the information will be applied.  

Words can be tricky things

We can’t know what someone really wants without checking. What you mean by effective, successful, better, ROI, or useful may not be what I mean. Plus, with acronyms and terminology unique to each profession, we are bound to misunderstand sometimes.

I witnessed an intense argument about whose fault it was that two weeks of effort was a complete waste of time, all because they interpreted the word “comparison”  differently.

(The business meant pre-versus-post. The analysts did Program A-versus-Program B).

Nobody clarified.

For these reasons, there is a critical need for effective communicators who facilitate productive dialog between business and analytic teams.

Essential communication skills for analytic translators are:

1. recognizing where communication goes off track
2. listening for signals that there is more to the story
3. asking effective questions to elicit different levels of information
4. selecting and delivering relevant information each team needs.

These skills are valuable, yet almost universally overlooked

In fact, these skills are so ignored in data-oriented professions that another search in Coursera for “communication skills for data analysts” did not return one actual course about communication.

The search listed about 75 courses, but none of them cover how to talk to other people.  A few do mention “language,” but those languages are programming languages. 

A similar search in Amazon does produce about 175 books, but only two are about the specific challenges of business-to-analytic communication.

Others are about data visualization, general listening and communication skills or completely unrelated topics.

If perceived importance is measured by the number of courses or books about a topic, these skills have not yet been recognized as essential. Which is a huge problem for the future of big data analytics.

On the other hand, for those who learn to do analytic translation well, it’s also a huge opportunity to stand out. 

Listen to this podcast on more about the power of listening, asking thoughtful questions, and getting to what really matters.

*note:  This is not unique to any one profession. It’s part of being human. All the more reason to work on ways to avoid it.