Hi everyone.

This is the middle of summer (for the Northern hemisphere to be precise), so I wanted to look at something that honoured the concept of the Summer Series, which, let me tell you, is a classic of French news magazines. 

I guess I’m counting on folks having both more time and a good chunk of their brain engaged in vastly more leisurely pursuits than reading about data-related topics. So I hope this strikes a balance of being engaging enough to justify the time staring at your tiny screen on a lounge chair, but not blood-boiling or worrisome like so many topics from our industry are these days.

All my best, Ariane

2 ways media companies are organising data

One of the topics that cross my inbox or Zoom calls with publishers touches on the question of the transformation journey for an organisation that wants to take data from a support function to a key function. Or even, the zero to 1 of the data team.

Two common presentation scenarios are:

• A: “We don’t have a data team to speak of and we want to know how much resources and time we need to have something valuable.”
• B: “We want to know how we compare to others our size in terms of how our team is built and how it works”

These are fascinating stories. And I’ll be honest, when I first started to consult for INMA earlier this year, I thought there were probably as many different configurations of data teams as stars in the sky. And that the stages of evolutions of these teams were probably quite diverse because no organisation really evolves quite like its neighbour.

Well, dear reader, I’m now eight months into this engagement, and I’ve learned that I was wrong. Publishers are really very diverse and they grow (or contract!) in their own different ways often enough. Their markets are so different, therefore their businesses are so different, too. Yet the story of how their data team evolved often resembles each other, and some of the milestones tend to be reliably present in each organisation’s history. 

What these stages of evolution tell us is how data gets metabolised in our organisation: First, as a nice to have that is there in a supporting capacity for specific functions, and, over time, as a driver of innovation and growth that touches many more corners of the company.

Capacity at each developmental stage of data function

I was working on a presentation for INMA’s North America board and trying to synthesize some of the stages of publishers’ data journey. Here’s what I came up with, understanding there are probably other ways to group organisations (very keen to hear how you’d swim lane this!).

Stage 1: Content analytics

Most publishers have some in-house support for analytics (data analyst or a fluent marketing person who will “run the numbers”). 

The data that is recorded is mostly around content and with simple product analytics that are recorded by default in tools like Google Analytics or Adobe.

In my observation, publishers at this stage do not have an established team called Data. The data function is wrapped in with a team that has the most use for the analytics function.

Stage 2: Subscriber data and customer-oriented business analytics modeling

My non-scientific survey is that most publishers in a subscription journey collect user data, and those that don’t have a CDP are considering it. The data department may still just be analysis, but this is the stage where publishers lean into more advanced business analytics to support their new or growing subscription business.

At this stage, there is a team called the Data team. It’s mostly analytics with a little bit of product management.

Stage 3: Product and user data, modeling, predictive analytics

More advanced publishers have evolved strategies for making their products “talk” with data engineering and data science to level up. Specialised data roles in advertising and CRM also are present. In my observation, the marking moment for when a publisher moves between Stage 2 and 3 is when they hire their first data scientist. I’ll explain why this is in fact a pretty big turning point.

When I chatted with Sanda Loncar, head of product and data at Kleine Zeitung in Austria, she confirmed this was very much their transformation journey, too. Sanda built the Data team at Kleine Zeitung. And for them, going from Stage 1 to Stage 3 took two years, building the team from one resource to five. 

What’s interesting is that the company quickly wanted more than what the team could deliver: “It was never about being able to get the budget,” Sanda told me. In other words, the transformation was organic — the team grew, adding new competences as the business asked more of the team. It is the opposite of transformation led by mandate or by external pressures like cost cutting or a deeply changing business landscape.

This model doesn’t apply to everyone

I am not addressing here the case of the digital natives — the Vox, BuzzFeeds of the world. This is because their data teams were usually there from day one, fully formed. When you’re born on the Internet, you’re going to do marketing and sales with quantitative data rather than with market surveys and user interviews (which is where most legacy media groups of course started their “data” journey — on the qualitative side of things).

So while the data teams of these organisations are interesting, they aren’t transformation journeys. For many publishers, the transformation is the tricky bit. An established data team is a great achievement. But once it’s established, their problems look like the problems of most other teams: how to scale, how to work cross-functionally, how to prioritise. 

I have also asked this type of question to data folks who are larger, multi-country organisations or groups that were born from successive acquisitions. These are interesting cases too, but their development looked different. By and large, the data functions were born earlier in certain parts of the org — say at a specific news property, or they were already present in a group that was being acquired) — and were either merged to become a group-level resource or used to be the foundation of a group-level resource. 

Let me know if you want me to tell these stories more in depth, but I’ll skip for now. These stories were more representative of the complexities of mergers and large group restructuring than they were about data teams specifically.

Stage 1 and 2: The birth of the data function and its growth

Data analytics in my observation is present at most companies — at least most INMA-member companies, though INMA’s membership skews towards the medium-sized org and bigger. 

The reality is analysis done from the modern tools we use (Google Analytics, Adobe, etc.) requires a good mind that is versed in understanding some of the concepts of traffic and funnels. But the tools have improved in their user experience, so you do not need to go through a six-month training to get going with GA.

That’s not to diminish the essential function of a data analyst (good ones are in demand!). But it explains why professionals with various backgrounds can find themselves in the role of a data analyst without necessarily having formally learned the trade.

So analysts can be folks who’ve become the resident expert of the organisation’s analytics platform, or they can be a more specialist role. And they are attached to whichever function first formulates the need for analysis: marketing, sales. It varies.

In the past few weeks, I have interviewed folks at a large publisher in Spain, a medium-sized publisher in Austria, and a regional publishing group in Germany to focus on their zero-to-one journey. All three organisations share the trait of being significant players in their market, mostly anchored in one country.

They also share the traits of being organisations that are now in Stage 3 — but did it using smarts rather than fat piles of cash doled out by their rich corporate parent (to be fair, all three organisations are also profitable businesses). As it says on the tin, we’re the Smart Data Initiative over here, and not the Rich Data Initiative. These three orgs were smart, first and foremost.

So going back to our more traditional publisher type (the national media group or single property publisher) — going from “There’s data somewhere in our org” to “We have a data team” usually came from a place of strategy.

Sanda Loncar, head of product and data at Kleine Zeitung in Austria, for example, said: “We knew we had enough of a need for data for Kleine Zeitung that just using a shared group resource was not enough.” She made the case to bring the existing analyst over in her product group (then called the Digital group) and form a proper team there. 

The Stage 2 of the data team seems to often get reached when analytics moves from being strictly a content engagement play to supporting a plan for direct reader revenue or complex advertising sales. A diverse group of constituents needs more dashboards, and the data team now has to scale itself. 

In addition, breaking down complex funnels and acquisitions means adding tools like Customer Manager Platforms and generally leveraging more first-party data. This is why Stage 2 seems to have more dedicated engineering resources for data than in the earlier phase — where work to support analytics wasn’t done specifically by an engineer specialising in data but instead by a front-end team following an analytics tagging plan.

Stage 3: From data that answers known questions, to data that tells a story we don’t know

I’d like to focus on what I think is a turning point in the maturity of data teams: the hiring of the first data science member of the team. All three of my interviewees articulated something that can best be described as: “We wanted to do more than what we found in our dashboards.”

The role of the analyst is to hear a question and frame a method to get an answer from a pile of numbers we gathered. The data scientist, on the other hand, looks for new information in the data (doesn’t come from a framed question necessarily) and attempts to see if predictions and a model can be driven from this data to uncover new truths we didn’t know about. (This whole post from Harvard Business School online contrasts these two functions.)

A thing that makes understanding both our content and our users tricky in our news publishing business is that even very vertical publications have diverse audiences. New articles and old articles get read. Navigation comes every which way. There are superfans and a bunch of drive-by readers we see once and never again. Or maybe we see them again but on another device and do we even recognise them? It’s a lot.

The reason I mention this is the resulting data we generate is also a lot. And the approach where we list all our questions about our audience and business and content could keep a large team occupied ‘til the end of time.

But the problem is, even if the question is interesting, there is no guarantee we will get an interesting answer. Data science, on the other hand, approaches the pile of data and says: “Let’s see what are the most interesting things we could find in there.”

This shift in mindset leads to that first data science hire. And the team there also reframes its role from being a service to the questions of other groups to taking the lead on bringing the company insights and predictions — letting the data tell stories about your users or business and giving answers to questions you were not asking.

Further afield on the wide, wide Web

• The Harvard Business Review looks at the evolution of the job market for data scientists, relative to when it declared it the “sexiest job of the 21st century” one year ago. It still a high-demand job (as you’ve no doubt experienced if you tried to hire), but the market has matured in a number of ways. 
• The Verge interviewed David Holz, the founder of Midjourney — an AI-driven image generator whose production you’ve probably encountered in your social media timeline. The interview goes from the scalability and costs of running a deep-learning model on, well, all the images from the Internet — as well as some of the issues at the bleeding edge of computer generated content, like copyright. It’s a great read. 

On deck for this summer

Later this month, we’ll take a look at the org chart question — where does data go? — and some paths taking in the scaling of the data team. And we will get to the other topics I mentioned earlier: AI-related projects for a bit of refreshing innovation — natural language processing and responsible AI, data science ethics. Write to me if you’d like to chat about these topics as I dig into them further.

Further up ahead: Jodie Hopperton, lead for the Product Initiative at INMA, and I have started working on programming our excellent Product & Data Summit in November. We’re digging into potential speakers with great tales of transformation and innovation. My inbox is open at ariane.bernard@inma.org if you have great ideas we should discuss!

About this newsletter

Today’s newsletter is written by Ariane Bernard, a Paris- and New York-based consultant who focuses on publishing utilities and data products, and is the CEO of a young incubated company, Helio.cloud. 

This newsletter is part of the INMA Smart Data Initiative. You can e-mail me at Ariane.Bernard@inma.org with thoughts, suggestions, and questions. Also, sign up to our Slack channel.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .