Skip To Content

Data Science Prose: Sharpening an Important Soft Skill

Reading time: 5 minutes
Data Science Prose

By James Cousins, Analyst Manager

I recently read through a mini data presentation from a friend of mine. Rather than simply answering a spur-of-the-moment question with a set of numbers, he chose to create a short slide-deck presentation conveying his findings. 

My friend was asked to figure out the current number of Coronavirus infections broken down by those who have and have not received the vaccine. He arrived at the number relatively quickly, not least because he’s been swimming in COVID-related inquiries lately. However, rather than simply say, “We’ve got n1 cases who were vaccinated, and n2 cases who were not vaccinated,” he chose to create a mini-presentation out of the ad hoc request. He wasn’t asked to provide context behind the numbers or to identify correlated points of interest; it was by his own volition that he took a little extra time to build a story around the numbers. 

It struck me as unique—a question small enough that it wasn’t part of an executive board meeting (or the like), but large enough that there were layers to peel back. That’s when it occurred to me: this is data science prose!

Data Prose

Figure 1: It goes without saying that you’ll find the answer to the questions you’re asked, and that you’ll deliver that answer to the stakeholders who need it. What remains to be decided is how you’ll deliver that answer. Keep in mind, too, that creating a mini-presentation doesn’t mean stakeholders have to see it. The act of treating a simple request like a presentation can be a solely intellectual exercise and still vastly improve your ability to tell the story behind data.

Why the Extra Mile?

As we develop our skill sets to include new methodologies, algorithms, and concepts, we generally agree that the data always needs to tell a story. However, in most discussions, the types of improvements are usually constrained to greater accuracy or precision in algorithms, enhanced data governance, or more efficient scripting. 

I suspect that most bootcamps and blog posts focus on techniques for analysis instead of communicating the results because techniques are commutable. Each technique may require specific conditions, but those requirements can be measured explicitly: k-means clustering analysis works as long as your variables are numeric, text analysis works as long as you have a free-entry text value, and so on.

The context required for compelling storytelling, though, is highly situational. Pundits and thought-leaders cannot spell out the ideal way to convey a story when it depends on the:

  • Urgency of the analysis
  • Preconceived expectations (from DS practitioners or stakeholders)
  • Results of the analysis
  • Localized implications of the results
  • Position and value proposition of your business within the landscape

The same results for two companies in identical industries could mean incredibly different things, which is why my friend’s data science prose struck me as groundbreaking.

How do you get to Carnegie Hall?

If you agree that storytelling and adoption is the real driver of value in your data science projects, and the tips surrounding data communication are sparse, how can you improve? What I mean to say is, how do you go from “know your audience” in the abstract to “share a data extract with Brigette before the meeting so that you can focus on the upshot in the first 5 minutes for Greg, then delve into the details for the general audience”? You get there the same way you get to Carnegie Hall: you practice!

Making a real connection to your audience is not a generic process. My recent experience helped me realize that you can avoid waiting until the big projects to practice your communication if you treat your smaller analysis like presentations, too. If you have to answer a simple, specific question about a recent performance indicator, take an extra moment to compare it against the previous period’s value. Then, consider seeing if you can find a correlate or two to help explain the change. Even if the presentation itself isn’t hugely impactful, the practice and the habit of unearthing the story behind the numbers—of writing the prose behind the data—will pay off.

With the constant ebb and flow of programming languages, pieces of software, and techniques, our core capacity to communicate the meaning behind data is a constant. Further, it’s the only part of our job that stakeholders actually require, because it’s the juncture at which our expertise meets their expertise.

Data Science Prose Presentation

“Thanks for coming to my TED Talk”

You’re all generally aware that you need to offer information in your stakeholders’ language and deliver it in a format they can quickly understand and act upon when presenting data-backed findings. What’s less apparent is that quick, small-scale presentations built on otherwise transactional analyses offer you a chance to test reporting techniques and experiment with visualization formats without risking a change to a larger, more critical report.

The truth is that we all need to make slightly more elaborate reports of our data from time to time. For many analysts in my professional circles, this is anxious-making because it’s a skill that usually comes out of the closet exclusively for high-stakes presentations: professional conferences, all-staff meetings, shareholders reports, and the like. This process of making mini-presentations—data science prose—will help you develop the skill before it becomes a necessity, which is a win for you and your organization.

Finding the Time for Data Science Prose Practice

This whole article has rested on a precarious assumption: that you have time to take on optional processes. Well, by design, these little forays into storytelling should be brief, so you probably do have enough time. You can have even more of it, though, if you can knock out the “numbers and figures” part of the process even faster. 

Rapid Insight’s tools and support can help you generate automated, repeatable reports so that you can get back to presenting analysis. The software’s visual data cleansing workflows are easy to build thanks to our drag-and-drop interface. You can generate reports of any length and depth to suit the need at hand.

Our support team loves meeting with customers to discuss ideas, share advice gained from other users, and review reports. This very article is living proof of our passion for translating our discoveries into quality of life improvements for our users. The one-to-one support model also makes training an easy and enjoyable process. You can get right to experimenting with how you can deliver information to your stakeholders in the most effective way possible.

Want to learn more? Click the button below for a personalized, one-on-one demo.


Notify of
Inline Feedbacks
View all comments