When delivering analysis, think like a newspaper editor

Analytics should be an editorial department, not an accounting team. Accounting delivers tables; editorial delivers stories. Accounting tells us what happened; Analytics tells us why it happened and what will happen next.

At least, that’s how Analytics should be. But often teams get pigeonholed as accountants, focusing solely on the What instead of the Why or the So What. The What alone is low value work.

Imagine a company preparing for IPO that issues an S-1 with only financial reporting tables. Or imagine a newspaper whose sports section only has box scores and standings, and whose business section only has stock prices and weather reports. Not that useful, right? Yet this is how many Analytics teams deliver their work. Readers get a flood of complicated charts and tables, or a deck packed with text, and are asked to figure out the story themselves.

What these numbers need to be useful is context

Explicit context is the easy part: Write up a description of what people are looking at and what it means.

“New signups are up this week 5%” has very little context. “New signups are up this week 5% because we increased marketing spend” is better. “New signups are up this week 5% because we increased spend in paid search by 10% as part of a test to start advertising in Germany. Early ROI suggests we should keep spending, but the long term impact won’t be known for at least six more weeks.” That’s better still. What would make it even better is a chart that immediately proves that the new signups are coming from German paid search. Give more explicit context when possible.

Context gets harder when you’re dealing with multiple objects. Now, in addition to explicit context, you’re providing context implicitly through hierarchy and adjacency and relative size and a bunch of other visual relationships.

In other words: You’re communicating context using the tools of graphic design, whether you like it or not. You can’t avoid this. Even a lack of design is design, because the reader will impose associations upon your work. Since ten charts can’t all be read at once, the reader will pick an order. The order they pick will affect the relationships they perceive between the charts.

The design decisions you make won’t just communicate order. They will communicate what’s important and what’s not, which questions are settled and which are uncertain, and what’s connected and what’s not. Good information design helps readers build the right narrative instead of the wrong one.

We can learn from newspapers’ design patterns

I worked for four years at a college newspaper. As a freshman I moonlighted on the Production & Design desk, laying out the paper late at night in InDesign. I really enjoyed the work: It was meticulous, time constrained, and visual. I didn’t realize at the time just how much the principles of newspaper design influenced my thinking. (Needless to say, though the work influence me it by no means made me a designer or a design expert.)

Newspapers have to deal with so many moving parts. The results aren’t always elegant, but they’re effective. I mean, just look at how much the front page of the NY Times communicates before you even get to the details of the stories:

You see hierarchy. You see order. You see connections being drawn between stories. You see hints at what you find deeper in the paper’s other sections. You even get a sense of the importance of a lead story relative to other days’ lead stories just based on the amount of real estate devoted to it. (You also see that most of the stories on the page are not straight news reporting, but interpretation — there’s something for Analytics teams to learn from that too.)

Each of the elements of the front page is governed by strict conventions. The lead story always sits in the top right. Bigger lead stories get more real estate. There’s always an editorial photo or chart above the fold. And so on. These conventions repeat day after day, reinforcing their power.

Often you don’t even notice the repetition until you spot an exception. Exceptions make it clear something important has happened. For instance, look at two front pages, just a day apart: September 11, 2001 and September 12, 2001. You don’t need to read a single word or look at a single image to know something important has changed. The structure screams: “Pay attention! The world is different from yesterday!”

Because I am an absolute nerd, I decided to abstract some front page designs in Figma to point out what’s going on. Here’s the basic structure of a NYT front page (I’ve simplified from six columns to five):

On a normal news day, you’ll see the lead story in the top right, a couple images, and a footer teasing other stories. Here’s January 3, 2022, for instance:

Here’s January 6, 2022. The all caps headline tells you something slightly exceptional is happening:

Here’s April 5, 2020. In addition to an all caps headline, the chart breaks through the NY Times logo. Something’s really amiss. The world is breaking in exceptional ways, so the design constraints of the front page start to break in exceptional ways too:

Here’s March 27, 2020. The all caps headline takes the whole page. And the chart is part of a block that rises from the bottom of the page to the top. This is a purposefully disorienting exception:

The last page, and one that still hits like a ton of bricks, is May 24, 2020. All black. Even the footer is gone, as if to say, “No other news in the world matters beyond these names.” Good lord. This is dataviz design at its absolute best, and it works way better than a chart ever could:

Design constrains process; process constrains information

These design conventions shape the story in profound ways. For decades the NY Times had a Page 1 Meeting with all the top brass. The constraints of Page 1 forced the team to decide what mattered and what did not. The design conventions of the paper guided how the organization ran, and how millions of Americans interpreted the world. They famously ended this meeting in 2015 because it didn’t match the continuous pace of online publishing. But the spirit of Page 1 lives on in the design of nytimes.com homepage:

You still see the same conventions, just digitized: Hierarchy, order, connections between stories, deeper links into the paper. (The paper Page 1 has evolved a ton too, as you can see here.)

Each newspaper has its own conventions. The NY Times puts the lead story top right. Others use top left. Tabloids take the whole front cover. (You can see a decent array of decisions here.)

Each medium has its own conventions too. Think about the structure of a magazine, with it’s cover, table of contents, articles, and often a back page feature. Or a scientific paper, with its title and abstract and body and citations. Or a movie poster, with it’s image and title and tagline. But for rich analysis, which often requires telling multiple related stories at once, I find the newspaper analogy works best.

When I’m communicating analysis, I think like a newspaper editor

The constraints of a newspaper front page are really helpful for clarifying priorities. What’s the lead story? How big is it? What else deserves to be above the fold on A1? What other stories do I need to tease on the front page? What other sections will I have beyond the front section? Within those, which stories are most important? Which deserve just a few paragraphs and which deserve a full feature?

These are exactly the questions you should be asking when communicating data analysis. I often imagine my work as a front page before I write up results:

The medium is the message

Of course, not every analysis is a sprawling story. Sometimes you have a quick ad hoc insight. Sometimes you have a recurring window into a stable part of the business. Sometimes you have a deep presentation on a single, narrow topic. Each situation has its own ideal communication medium.

And the medium is the message. Think hard about what the appropriate medium is for communicating insights. How they are delivered will tell the recipient as much about the analysis as the analysis itself. Ask yourself, “If this analysis were physical, what would it be? Would it be a newspaper? Would it be a movie poster? Would it be a three act play? Would it be an ad on the side of a bus?”

There’s no simple mapping here, but instead a question of daily judgment. It’s up to you, the editor.