In this issue:

  • Birthday intro
  • How can we get to better digital data? Part two of our exploration of data science and publishing
  • Links of the week: Social norms and sound towns
  • Did you read? A preview of our 2024 initiative

One of the cool things about having a birthday in mid-December is that I have an excuse to commence year-end navel-gazing a little bit early. I get a head start on contemplating resolutions and personal goals.

Anyway, I celebrated my birthday on Monday. Stared at my belly-button. Made lists. Thought about wins and losses. I'll share some next week.

In the meantime, let's reflect on the end of an era. I have not read Ben Smith's Traffic yet, but in an effort to take read more industry books in my 41st year, I've put it on hold at the library. (I also put a hold on Steven Soderbergh's 2002 film Traffic because why not? Same keyword, great film!)


Sponsors and promotions

Learn sustainable, algorithm-proof content strategy

Join our 2024 cohort to learn methodologies and measurement that improves digital experiences and boosts your business

Explore the cohort

When it comes to content, is data science actually scientific?
Understanding predictive modeling can help you play moneyball temporarily, but it doesn’t scale as anticipated. And the scientific method completely falls apart when applied to the chaos of the internet. Here’s why.

Read Part I of this essay

From engineered content to informed editorial: Blazing new trails in digital publishing

by Deborah Carver

Fatalism can be attractive in industries that manufacture heroes, villains, and conflict to promote a tight, uncomplicated story. With data science and algorithms being panned into binary, good-versus-bad narratives, where can we find clarity and room to experiment?


Working online in the Buzzfeed era

I'll admit that I talked up Buzzfeed as much as any pundit in the early 2010s. The new digital darlings were more upbeat and generalized than the folks in the Gawker network. They were innovating with digital tools and interaction, especially in their quiz department, and hiring talent whose work was infinitely sharable. But despite The Dress, I always wondered how exactly they were making their money. 

As a content strategist, my job was to connect content performance with leads generated for clients. Generating high-quality leads with content smoothed out my clients' sales process. It was direct-response marketing at its most basic, built in the exact model of the direct mail industry.

My job was to look at client data directly and figure out how my content and SEO suggestions impacted performance. I used the analysis skills I'd learned in grad school and on the job to make reports and provide proof that, yes, content optimization made an impact. The economics of sending qualified searchers to websites with retargeting pixels made sense from a practical perspective, even if the practice creeped audiences out.

We saw the traffic, but where was the revenue?

Based on what I knew about the internet, analytics, and my clients' websites, I never really knew how Buzzfeed was making money. Buzzfeed didn't have ads running on their site in those early years, meaning that they weren't directly monetizing the traffic they attracted. Sure, they attracted gargantuan amounts of traffic from social, collected oodles of irrelevant personal data via quizzes, and probably hosted retargeting pixels on their own content. 

In between sharing quizzes and arguing over The Dress, my coworkers and I wondered, aside from venture capital, how did Buzzfeed make money? At the time I figured that they knew how to run a digital content business simply because they were doing it, but I couldn’t see the mechanics of profitability on the face of the product. 

No one profited from retargeting pixels except Facebook, Google, and the advertisers they served. And perhaps Buzzfeed sold predictive profiles of their users back to data brokers, but at the end of the day, knowing which type of cheese a bunch of bored millennials identified with isn't considered valuable data outside of the dairy industry. What matters in corporate data science aligns directly with purchases, which makes credit card companies, not publishers, holders of all the good stuff.

Meanwhile, after shifting to digital marketing, I wasn't allowed to make a single creative suggestion to a client without some kind of data to back it up. "Gut" and "experience" weren't acceptable as explanations, even though they had been in the publishing industry. My stakeholders required numerical proof, either gathered from the broader internet or one of the analytics tools I used, that what I suggested would work. What passed for editorial judgment at previous jobs was disregarded entirely in the days of growth hacking.* 

*I cried a lot when I was a junior strategist.

When experience in digital business contradicts the Buzzfeed narrative

I always wondered whether Buzzfeed's executives received the same kind of scrutiny that my early efforts at my first digital strategy job did. Were their bosses asking them to prove how pageviews meant business? Did they have to draw a direct correlation between audience size and revenue generation? Buzzfeed’s success certainly contradicted my experience, which was that from the early 2010s until now, large amounts of traffic — no matter how charming the content — never indicated that the business was successful.

Later, when I argued with publishers that, based on the data I'd seen with my business clients, I didn't see that digital audience size directly correlated with revenue or business performance in any way, those suppositions were waved away. Obviously the publishing industry knew what they were doing with the tech they were yielding. Clearly Buzzfeed and its ilk knew how to use search and social to their advantage as a business and were using the data they attracted scientifically and correctly. Otherwise they wouldn't be doing it

I learned that I was best keeping my experience and findings to my business clients, and the work I was doing was well-paid, so I had no reason to complain. 

Now we're at the end of the worst year in digital media history, per Adweek's Mark Stenberg. Media pundits — who write daily newsletters but don't analyze publisher performance directly — cry out at the decline of online publishing. Layoffs are everywhere in most professions described as "knowledge-class," especially that knowledge class dealing with content and language.

Most websites from legacy publishers are overloaded with clunky programmatic advertisements. Buzzfeed now runs ads on its sad user-generated content, making its audience a fraction of what it once was.

Avoiding fatalism and embracing evidence: The publishing-data divide

Unfortunately, some smart people in the publishing industry have taken Buzzfeed's business missteps to indicate that all data is bad. Instead of reexamining the assumptions of Buzzfeed's dominance and the clear mistakes of management to connect audiences with sustainable revenue, they point to its model as an example that "data" never worked to begin with, whatever that means.

Or they're subscribed to the belief that we are at the end of an era. That the internet will always be this shitty. That ad tech is as good as it will ever be. That the men in charge have always known exactly what they are doing. That we all just have to resign ourselves to reading website content on a small sliver 15% the size of the total smartphone screen. Some have given up on algorithmic distribution entirely, despite the data that says search still works, quite well, to attract engaged audiences. 

Fatalism can be attractive in publishing, an industry that manufactures heroes, villains, and conflict in the service of a tight, uncomplicated story. It's enticing to bait human cognitive biases and stoke conflict to boost engagement and traffic. Algorithms: bad! Publishers: good! Data: bad! Gut: good!

And yet great content-focused websites still exist, built by publishers who realize that data is important in digital production, as long as it's used appropriately.

The two conflicting interpretations of digital publishing in 2023 — that data science is an unfuckwithable predictor of success, or that all the algorithms are bad for digital content, always — tell an easy story that chills what should be a period of actual experimentation. 

So, based on the evidence in front of us, how should we proceed?

1. Separate the good data from the bad.

Just because something is positioned to be scientific doesn't mean that it’s infallible. But it doesn’t mean that content professionals should ignore or wave away the evidence that our owned audience provides.

In content strategy and editorial planning, digital behavioral data can be extremely helpful when it's used to understand an audience. We can use search data to understand what people are interested in and how they think about certain subjects. 

We can use our owned website data to understand what content engages readers but doesn't attract algorithmic signals. We can take the inputs from digital data sources to organize our content with architectures and structured data that make sense to both algorithms and people. And we can ask our audiences directly for feedback, should they want to provide it.

But we have to keep looking, benchmarking, and figuring out what works for our audiences in a nuanced way.

2. Remember that audiences don't behave monolithically.

One of the first rules of user research is to resist the individualist assumption that your audience behaves as you do. Avoid a focus group of one, researchers say. Because if you've ever conducted user testing or looked over a relative's shoulder as they use the phone at the holidays, you understand that people use digital media in wildly, shockingly different ways.

That matches with one of the first rules of the publishing business, which is that being well-read or well-spoken doesn't make you an expert in publishing. Spending time online doesn't mean you know which content will be successful from a business perspective. No matter how much you get cross-eyed from "the discourse" or think you understand what online audiences know from the front-end of the internet, you're probably missing the mark.

Researching an audience isn't new to digital media. It's been the job of commissioning editors for decades: before choosing a content direction, publishers want to know whether a particular concept will be successful. This means conducting interviews, examining your owned data, and understanding how audiences use digital content. Some people call it "human-centered design"; others call it "due diligence." I call it evidence-informed editorial. 

3. Triangulate sources before drawing a conclusion.

As editors and creatives, we need to recognize that digital data signals are good inputs and can directionally lead us to new ideas or away from risky ones. But these data neither predict nor determine what will resonate with the audiences who want to read, watch, or buy from us regularly.

Using a variety of sources will guide you toward better editorial decision-making and more valuable content. These directional sources should include a selection of the following:

  • In-person interviews with subject matter experts who are responsible for outcomes
  • Informed analysis of your owned analytics (and not automated "insights" without additional human scrutiny) 
  • Guided textual analysis of customer service records, social network chatter, and reviews
  • Organized and prioritized keyword data (but not trends data)
  • Audience surveys
  • Longitudinal research conducted by expert institutions — such as Pew or NORC

Single-source stories are neither informative nor engaging. Focusing on individual anecdata — a lawsuit here, a bad experience there — may attract viral pageviews from the extremely online crowd seeking novelty and gossip, but won't support your content efforts long-term.

4. Churn is natural. Embrace it.

In the service of pleasing as many people as possible and raking in cash while building an audience, publishers and technology companies alike have a strong corrective impulse to extrapolate limited findings of a new opportunity to a large amount of people, fix the identified "problem" as quickly as possible, and check it off the team's to-do list.

But often that means frustrating an audience that prefers consistency over experimentation.**

The to-do approach to both publishing and software development is as hamhanded and reactive, likely to anger existing lurkers who were happy with the way things were, and frustrating to a creative team who wants to build something new. We have to remember that building online businesses is a social activity rife with change, even if we don't see the satisfied customers as prominently as the angry ones.

But we don't need to react to every individual's opinion, every trending topic, or every market shift as if it were a mandate. Most people just want a stable build that meets the expectations they had at sign-up.

**Hello Content Technologist audience, I realize this year has been inconsistent! And I see you. Expect more details and consistency in the next few weeks.

5. Welcome research findings as a creative starting point, not a limiter.

There's a feeling in digital media, especially for those who are new to the industry, that we've already run the gambit of creative possibilities online. That Buzzfeed did all the experimenting forever, and now we're all stuck with this terrible reality where digital media is irrevocably tarnished and everything is hopeless.

But realize that we're just at the beginning of the digital media age. Just because there is one Wikipedia and one doesn't mean there can't be a second and a third.

When you have done your homework and landed on some really good insights — whether you've found hidden goldmines in keyword research or identified an unmet need in audience interviews — avoid the impulse to take those findings literally. Instead, chew on them, consider their contexts, and let them inform future decisions.

Digital publishing is a creative industry in its early stages. Only certain behaviors can be recorded digitally; the data available to us is by no means exhaustive; and culture is a conversation that largely occurs offline. So let’s bring more data to the table and make editorial decisions in an informed, intentional, researched manner — without thinking that our process needs to be perfectly scientific, or that we’ve already determined the most empirical direction. 

This year may have been a downturn, but I have to believe, when I look around at all the paths yet to be forged, that online publishing is just getting started.


The Content Technologist a company based in Minneapolis, Los Angeles, and around the world. We publish weekly and monthly newsletters.

  • Publisher: Deborah Carver
  • Managing editor: Wyatt Coday

Collaborate | Manage your subscription

Affiliate referrals: Ghost publishing system | Bonsai contract/invoicing | The Sample newsletter exchange referral | Writer AI Writing Assistant

Did you read? is the assorted content at the very bottom of the email. Cultural recommendations, off-kilter thoughts, and quotes from foundational works of media theory we first read in college — all fair game for this section.

More next week, but here's a sneak peek at an experiment I'm working on for next year.