Features Hub

10 data storytelling mistakes to avoid

Mon 18 Nov 2019 | Nate Nichols

Nate Nichols is distinguished principal at Narrative Science, an AI-powered software startup that turns huge amounts of data into plain-English stories. We asked Nate to explain why enterprise data storytelling often fails to make the grade

We live in a data-saturated world. But data only matters if you can turn information into action. To get from one to the other, you need to tell a story about your data that any person at your company can understand.

Why? Because numbers and charts don’t motivate people to seek better outcomes. Stories do.

So how do you tell these stories?

Here are ten mistakes to avoid in order to tell a good story (about your data) that will drive strategic business.

Not having a hook

Imagine arriving at a data rundown meeting and the presenter clicks to the first slide and says, “If you’re wondering what these numbers are saying, let me spare you the time: not much without some context! So let’s jump into that context and the story behind the data.”

This uncommon opener – and leading with an anecdote – are each examples of good hooks. Ideally, this description caught your attention and compelled you to keep reading.

If you’re presenting to executives and key company decision makers, you’ll want to lead with a memorable title or opener: something unexpected, something funny, or something intriguing that gets at your main idea. Tell them from the start why they should care.

Not setting the scene

Data without context means nothing. If you only say, “We got five sales this quarter,” you’ll leave your audience wondering about the size of those sales, how they compare to previous quarters, and whether they met expectations.

Your headline is not the story. Don’t just say, “Sales increased 12 percent this month.” Tell your audience why that’s important because of how sales numbers have changed over time. Try instead, “Sales were stagnating for three months, and this month they began to rebound, increasing by 12 percent from last month.”

By phrasing it this way, the sales team can reflect on what they started doing differently last month and do more of that, rather than just being happy that there are more sales.

Not using enough words

Numbers are essential. But often, numerical information can be distilled into themes and trends – in other words, into words. Staggeringly high or low percentages can show impact, but consider verbs that convey the same meaning: “peak,” “level off,” or “rebound.” If you have a regression model, for instance, describe “how likely” something is.

And don’t rely too heavily on basic statistics words like “significance” or “normal” that have broad meanings in a non-data context. Who knows when members of your audience last took a stats class – whenever possible, convey meaning in plain English.

Not telling enough stories

A robust data storytelling program means that everyone stays up to date on key insights.

A quarterly roundup is too infrequent: chances are there will be too many data happenings to include. Either you’ll overload your audience, or you’ll skim over important details for the sake of time.

In an ideal world, employees have access to a quick, couple-sentence summary of data news every day. But data scientists are busy, and manually putting together stories every day is unrealistic. Aim for a weekly or monthly schedule for sharing data stories with the rest of the company.

Only telling stories about big news

Establishing a regular data storytelling cadence means that you’re not always going to have big news to report. And that’s okay.

Big changes or insights are great, but often the data just isn’t that dramatic. A lack of change is still data – and still a story worth sharing. Your colleagues need to understand what your data is telling you about business at all times.

If numbers for a marketing campaign have been looking the same week to week, you could say there’s “nothing new to report,” but instead you could say that “despite the launch of two different email marketing campaigns, we’ve had no new inbound leads.” The latter contextualises the work behind the data so that your audience can make decisions about what they can do differently.

Including unnecessary details

You’re not obligated to include everything. Fight the impulse to include details that you find interesting or worked hard for but that ultimately don’t serve your big idea. Or save those details for another story.

Time is valuable for everyone, so your story should reflect only the details that relate directly to business decisions.

Letting data visualisation speak for itself

Data storytelling education often focuses on how to draw attention to the right bar in the chart or how to label your axes appropriately. This is good data visualisation, but it can’t tell the whole story.

Data storytelling shouldn’t be the voiceover you offer over a slide. Data visualisation is a great supplement, but charts, numbers, and graphs don’t always speak for themselves – just as a picture book needs words to guide its story.

Don’t assume that visual representations of data are digestible to every non-data scientist. Start with the story, and bring in visualisations when they can illustrate a key point.

Thinking of stories as “dumbing it down”

Dealing with information complexities is a core part of a data scientist’s job. Encapsulating those complexities in simple language, then, might seem like a reduction.

It’s not – it’s just a method of communication. Data storytelling isn’t watering down hard-earned data discoveries; it’s translating them into a language that everybody speaks. Stories don’t dumb data down, they make it smarter.

Closing the conversation after the presentation ends

Once you’ve told your story, start a dialog. Open up the floor for questions. Start talking about action items.

Telling a story means fewer questions about how to interpret the data and more questions about how your colleagues can use this information to make decisions. A good data story should be a discussion starter.

Treating data storytelling as a nice-to-have

Data stories are nice to have, but as we move forward to unprecedented levels of data, we need to sift through to pull out what matters. Without stories to make sense of it all, your next business breakthrough could be lost in a sea of charts. Data storytelling has made the transition from nice-to-have to need-to-have.

Share Data Stories with Everyone

Business works best when it functions like an organism: every team communicates with one another through a shared information network. If data is your lifeblood, then data stories are its heartbeat. Data scientists who see the most success in business will be those who are able to efficiently circulate data-fueled stories to all of a company’s essential organs.

Experts featured:

Nate Nichols

Distinguished Principal
Narrative Science


data-driven visualisation
Send us a correction Send us a news tip