Mastering audience segmentation with Selma Stern

Why prioritize segmentation?

Segmentation can be mind-bogglingly complex and wildly impactful.

The word comes from biology, where it refers to the process of cell division from a fertilized egg to a full organism. In business, it refers to identifying the constituent parts of a whole – ideally, a set of mutually exclusive and collectively exhaustive groups of customers. It can’t be perfect by definition, but even an imperfect segmentation can massively improve your bottom line.

At its core, segmentation addresses three fundamental business needs:

• Stickier content: By clearly defining  Ideal Customer Profiles (ICPs), newsrooms learn not just what to write about, but how to speak to their readers so the content truly resonates. If you replicate your ICPs as agents and train your newsroom to use them, you can get endless customer feedback at any time of the day.
• Smarter monetisation: Without segmentation, a dynamic paywall is only based on recency and frequency – which is fine, but does not capture your full market potential. . With it, you are “fishing where the fish are,” targeting the right users at the right time.
• Better marketing: Different segments live on different channels and respond to unique triggers. Segmentation tells you exactly where to find your users and helps advertising brands connect with theirs.

Choose your lens:

To build a complete picture of a user, like “Amanda, 40, VP of Marketing,” organizations can look through several analytical lenses:

Segmentation type	Focus/use case	Signal examples
Behavioral	UX Optimization	>5 visits per month, scroll depth, newsletter opens
Needs-based	Product Development	Prefers summaries, explainers, or opinion pieces
Value-based	Revenue Optimization	LTV, subscription tier, payment history
Technographic	Paywall Optimization	Device type (Mobile vs. Desktop), OS, Referral (LinkedIn vs. Facebook)
Demographic	Ad Sales / Strategy	Age cohort, geography, seniority

> Spektrum’s evolution of audience segmentation and testing

Five tips for succeeding in segmentation

1 – Hire strong analytics: This is the “single most important unlock.” You need talent that can bridge the gap between high-level business strategy and deep data modeling.

2 – Make it cross-functional: Segmentation is a “Left Brain + Right Brain” exercise. Editorial, product, marketing, and data teams must own the model together; no single department can succeed in a silo.

3 – Quantify benefits early: You must build a business case or risk losing organizational priority. Use “Napkin Math” to show the potential uplift: for example, if a newsletter with a defined audience grows twice as fast as a vague one, what could the vague one be worth with a properly defined audience?

4 – Remember the customer: Segments should feel like people, not rows in a spreadsheet. Give them names like “Jane, the CMO” or “Joe, the Sales Lead.” Talk to them, build agents around their profiles, and ask, “Would Jane actually read this?”.

5 – Keep it simple (but useful): Aim for 4–6 mutually exclusive, collectively exhaustive (MECE) segments. If your team can’t name the segments from memory, your model is too complex and will fail to be actionable. Even if your goal is to build a dynamic paywall algo – lose (human) control of complexity, and you won’t know what’s working, which means you won’t be able to adjust next time the world changes.

Real-world impact: fewer paywalls, more subscription revenue

A US publisher’s journey from a rigid freemium model to smart segmentation highlights the power of behavioral data. They discovered that just 1% of traffic drove 25% of conversions.

By analyzing attributes, they found that newsletter click-throughs and direct repeat traffic were high-value, while Facebook and Android traffic converted at near-zero rates. By stopping the paywall for near-zero segments and doubling down on high-value behavioral and technographic signals, they achieved:

• A 2x increase in conversion rate.
• 50% fewer paywall exposures, leading to a better user experience for the same number of new subscribers.

Did this publisher get it completely, fully right? Absolutely not. But the model was a dramatic improvement over the simplistic frequency and recency model they were using before.

As statistician George Box famously said: “All models are wrong, but some are useful”. Be pragmatic—start with the data you have, build empathy for your readers, and iterate until your segments drive measurable growth.