Step 1: Cutting complexity at the paywall 

The first meant shortening the time between “I want to read this article” and checkout. Removing the higher-priced products from the initial paywall felt risky and seemed, on paper, like we’d be hiding potential revenue. But we did it anyway, betting that getting users into the product experience fast mattered more than upselling them right at the door.  

By eliminating the need to choose between several packages, we initially achieved a significantly higher conversion rate. 

Our new default paywall offer became H+ Standard digital access. In addition to a trial,  we offered users the option to commit to a three-month subscription at a 30% discount  compared with the monthly price. 

The effect became apparent quickly. With no package comparison to work through,  conversion rose significantly. Communicating less information turned out to be the clearer message and simplifying the offer helped remove friction. 

Step 2: Moving from a one-size-fits-all paywall to propensity segmentation 

During our tests, we observed that H+ Premium was also in strong demand. The product clearly appealed to a high value group of users.  

So, we reviewed the paywall flow again with a different question: how do we keep the simplicity of a single-offer paywall without showing every reader the same single offer. 

We started by splitting our audience into two groups: 

• Users with a low likelihood to subscribe 
• Users with a high likelihood to subscribe

Our hypothesis was that users in the high-likelihood segment might have both a higher willingness to pay more and genuine use for the extra Premium features. To test this assumption, we used our Likelihood-to-Subscribe (LTS) score to decide which users would see the Premium offer. The central question was where to set the cut-off and how to present Premium without smuggling complexity back in. 

Step 3: Optimizing for revenue, not just conversion 

Finding the right cut-off took more than looking at the initial conversion rate comparison. We tested different thresholds and evaluated several downstream metrics: 

• Initial trial conversion 
• Trial to paid conversion 
• Resulting subscription revenue 
• Retention and subscription lifetime  

This broader perspective was essential because H+ Premium generally converted less reliably from trial to full price than H+ Standard. So, a paywall variant that wins on initial conversions was not automatically the better commercial option and can still lose on revenue. Premium trials that never convert to paid don’t just underdeliver, they replace the potential Standard subscriber that we would otherwise have kept.  

Our analysis therefore focused on the entire revenue stream rather than on the initial  conversion alone. The decisive question was therefore never “which offer converts  best?” but:  

“Which offer generates the highest sustainable revenue per user segment?” 

What we have learned

Our experience showed that paywall optimization is not necessarily about presenting more choice. In our case, offering less choice raised conversion and segmentation then allowed us to reintroduce a higher-priced product selectively, targeting users who were more likely to value and pay for it. 

The result is a more differentiated paywall strategy: 

• A simple, low-friction entry for users with a lower likelihood to subscribe
• A higher-value Premium offer for users with a stronger likelihood to subscribe
• A measurement framework that considered both conversion and downstream  revenue 

The most important change was moving away from a one-size-fits-all paywall. Instead of asking every user to choose between several products, we used user-level signals to decide which offer was most appropriate. 

For us, ARPU growth did not start with raising prices. It started with reducing complexity, understanding user intent and matching the right product to the right audience. 

About the authors 

Ann-Kathrin Wind is a Performance Marketing Manager in Marketing and Sales at  Handelsblatt Media Group. 

Marvin Suder is Topic Owner Conversion in Product and Technology at Handelsblatt  Media Group. 

Together they work in a cross-functional team on onsite optimization for Handelsblatt  and WirtschaftsWoche, from paywall design and audience segmentation to conversion  optimization. 

About Handelsblatt Media Group 

Handelsblatt Media Group is Germany’s leading publishing house for business and  financial journalism. With brands including Handelsblatt and WirtschaftsWoche, a  growing portfolio of digital information products and a strong live events business, the  company delivers independent, reliable business intelligence to decisionmakers in  business, politics and society.

“What you are trying to do is to put the conversion (from free user to paying subscriber) into the user journey as soon as possible. Because that’s what you want… By having a nice “beware, premium content ahead” icon (or, should we say warning), you are short circuiting that thought process and user journey. Conversions need to happen based on the article on your website, not how you market it in a newsletter or on the homepage.”

It’s perhaps one of the reasons why we’ve seen increasingly more publishers trying the opposite, i.e. signalling when content is free to access.

The New York Times and The Verge for instance have been testing this for a while now, dedicating a whole section of their homepage to “free to read” content.

The Independent in the UK also has a tag on a newsletter promotion, highlighting the 1 month free.

Subscription without blocking content

Most publishers solve the reader-revenue question by drawing a line through their content. Some of it is free to access, some of it is paywalled. SFGATE deliberately didn’t do that. The web stays entirely free, exactly as it’s always been. What’s new sits next to it, a separate, paid app.

This means that the web traffic and engagement on-site is preserved whilst the brand also monetises their most engaged, loyal readers through an additional product.

> Read the full story of SFGATE’s subscription launch

Subscriptions focused on community and relationship-building

In an age of content abundance, where generic content becomes easier to produce, the advantage shifts to what is harder to replicate: original journalism, trusted relationships and communities that give audiences a reason to return. It’s what FT Strategies has labelled as the Community Era, and it’s exactly what many publishers are moving towards with subscription products built for lasting relationships.

The Economist launched Play, a $15/month audio and video-only tier that sits inside the existing Economist app. It sits alongside a second, deeper relationship-building product: The Economist Insider, a premium video series featuring the editor-in-chief and senior journalists interviewing figures like Christine Lagarde and Mark Carney. A year after launch, 75% of subscribers have engaged with Insider at least once, and churn drops further the more deeply someone engages with it, holding even through a recent price increase.

Together, Play and Insider represent two ends of the same strategy: Play is built to start a relationship with a reader who isn’t ready to commit to the full product yet, Insider is built to deepen the relationship with a subscriber who already has.

WIRED also relaunched their subscription offer in Summer 2025, priced at $48 a year (roughly $4/month) with a focus on connecting “our humans to all of you humans.” (Katie Drummond, WIRED):

• A five-newsletter subscriber suite. Two entirely new verticals launched alongside three previously free, popular newsletters moved fully behind the paywall. All five are written conversationally by senior correspondents and built to work as standalone “inbox reads,” not link round-ups, which is what makes them habit-forming rather than skimmable.
• Structured, cadenced livestream AMAs, running at least twice a month, with a staff moderator and two senior beat reporters per session. The tight panel format, rather than sporadic ad-hoc video, self-selects WIRED’s most loyal readers and puts them in direct, live conversation with the reporters covering their beat.
• Expanded commenting, rolled out across a much wider set of articles than before, specifically to turn passive readers into visible, active participants in the conversation around a story.
• Subscriber-exclusive narrated articles, giving the subscription a genuine audio dimension rather than treating audio as an afterthought.

Alternative ways to bridge the gap to subscription

Anjali Iyer, Global Head of Subscriptions at The Washington Post, delivered a session at the Audiencers’ Festival in London focused on how news organisations must adapt to modern consumer habits by expanding beyond rigid subscription structures.

Anjali shared how data exposes a massive gap between the willingness to pay and the willingness to commit. Modern readers, particularly younger cohorts, operate in a fragmented digital landscape shaped by streaming services, search loops, and social algorithms. They demand short-term choices and are often willing to pay a premium price for temporary access over a recurring subscription package.

The Washington Post’s solution is flexible access models. By testing micropayments (weekly and pay passes, as well as pay-per-article with an in-paywall mobile payment module), they’re uncovering an untapped market: 42% of flexible access buyers are brand-new, anonymous users outside the existing ecosystem.

Interestingly, this also serves as a nurturing mechanism for long-term subscriber growth. Over a 180-day period, up to 8% of week-pass buyers, 4% of day pass buyers, and 3% of pay-per-article buyers eventually became full subscribers.

> Read the full story from The Washington Post

A handful of publishers are also testing “pay what you can” subscription models.

For instance, Canada’s Les Coops de l’information, a cooperative of six Quebec news outlets, let readers name their own price for the first three months.

Against a C$1 minimum, the average paid was C$3.12 a month, and some new subscribers chose more than the regular subscription price. The suggested figure in the price field did the work – suggesting C$5 lifted the average to C$3.39 and suggesting C$6 lifted it to C$3.50, with more people simply taking the suggestion each time.

84% of the acquisitions were from readers who had never subscribed before, which is the number that matters most for a local publisher wondering whether it has exhausted its market.

> Read more about Les Coops’ pay what you can campaign

Selection over unlimited access

Faced with information fatigue, unlimited access as a subscription product is not always the answer. Les Echos is instead making the opposite bet, that of selection. “La Sélection” consists of between 9 and 15 articles a day, chosen from a daily output of 120 to 150 pieces, following a reader-needs grid (inform, reflect, get inspired, develop, unwind) and a strict quota discipline.

Three platforms (a dedicated tab in the app, a section on the site, a daily newsletter), a weekly long read on Saturdays, and 5 shareable articles a month. Early bird price sits at €9.90 a month.

> Full article on Les Echos’ “La Sélection” here

Not trying to sell on day 1

When benchmarking 100 subscription businesses, including the NYT, Figma, Miro, Calm and Masterclass, we noticed that conversion almost never happens in one step. It happens through micro-conversions, each with its own objective. And more importantly, even conversion itself seems secondary. The real priority for those companies is to create value for their user as a first goal. To build engagement. To make the product feel indispensable before asking for money.

The critical principle behind this model: each step has one single goal.

Figma: $12.5B company.
Homepage: “Get started for free.”

Then the onboarding starts: What’s your name? How do you plan to use Figma? What do you do for work? Where do you work? Have you used Figma before? What do you want to make first?

Six questions before you see anything about pricing. And when the plan page finally appears, it includes a free Starter option, so even the pricing step isn’t a hard sell.

The New York Times even tested this earlier in the year. The only action button in the top right corner (on the left) says “LOG IN”, because at this stage, the goal is registration, not subscription.

Countdowns

Whether on the paywall, subscription offer page or in the registration onboarding journey, publishers are using countdowns to create urgency and encourage conversion. It’s a small change but one that’s proven to convert.

> Find 100+ paywall benchmarks here

Code is now accessible to anyone with a laptop and an internet connection. The old bottleneck, engineering resources, is gone. That should mean launching more ideas, iterating faster, and learning about our audiences and our mistakes along the way.

At the Minnesota Star Tribune, where I launched our first two audience-facing AI products, the most useful lessons came from how readers responded, including when the products missed the targets we’d set for them.

So how do you hold the line on credibility while building products you’re willing to let fail? A few suggestions:

Build a public sandbox

Separate the experiment from the main product. A subdomain with a clear label (beta, labs, sandbox) tells readers what they’ve found and resets the expectations they’d bring to the homepage. The readers who show up are opting in, and they still give you real behavior data. From there, the experiments that work can graduate to the main site.

At the Star Tribune, our first audience-facing AI product came out of a four-week experimental sprint. The result was Culinary Compass, a restaurant guide based on a reader’s in-the-moment vibe. We launched it on the primary domain, and readers who found it there expected a Star Tribune restaurant guide, not a rapid experiment. It told us a lot about what our audience expected out of us, and what they expected our use of AI to deliver for them. A sandbox would have taught us the same thing but for less.

Give it an end date before it starts

Decide when the experiment ends before it launches. Better yet, experiment with something that has a natural beginning and end. This helps remove the call of pulling something, so if the end date was always the plan, no one has to make the call to shut it down.

I launched Fair Bot in 2025 for the Minnesota State Fair knowing it would only be useful for the 12 days the fair ran. It let readers use semantic search to explore more than 300 food vendors by craving instead of by name, for example “something fresh” in a sea of fried food, or “sweet but not too salty.”

The tool logged more than 10,000 searches, but a surprising number of them asked where the bathrooms were. We had built a food finder, but our audience wanted a better fair guide. When the fair closed for the year, so did the bot, and the insights we gleaned went into planning for the next one.

Show your work

I miss the blogs newsroom product teams used to keep about how they built something. A lot of thinking goes into a product decision, and most of it disappears once it ships. Writing it down forces the team to say what they believed and why, and it gives the rest of the industry something to build on.

I worked at Vox Media from 2015 to 2022 in roles that straddled the newsroom and product teams, and I had the chance to contribute to a few posts for the Vox Product blog. I was reading the blog before I worked there, and it shaped how I thought about the job long before I was in the seat.

Product blogs are also where a version of a post-mortem can live. We already have a practice for when we need to issue a correction, and a flavor of this practice should extend to our experimental products.

We work in an industry that expects transparency from the people and institutions we cover. Holding the products we build to some version of that standard is a reasonable place to start, and it’s the part of the journalistic mental model that experimental products should look to keep.

Let’s compare notes

I’m the AI Strategy and Product Advisor at Hacks/Hackers, and I want to take the lessons above to independent newsrooms, the ones without a product team to spare. I work with news organizations as they think through where AI belongs in their work, from adoption and training to tooling and audience-facing products. Whatever the project, the work starts the same way: try it, watch what happens, and write down what you learned.

“We didn’t look at this as let’s substitute one for another. Why would you? That’s leaving cash on the table,” Burney says. Print subscribers are highly profitable, the volume decline is gentle, and their engagement is incredible. So, the apps were sold to people who had never bought a magazine. “What we’ve done is create a new audience which previously did not exist. So our growth has been genuine, rather than simply substituting one channel for another.”

• The Good Food app came first, launching around 2022. It now sits just short of 200,000 paid subscribers, from a standing start
• Nutracheck, a diet and nutrition tracking app, was acquired in 2023 and cross-sold to the Good Food audience. It is profitable, has roughly doubled since then, and currently has 580,000 subscribers
• Radio Times followed with its own paid products, including What to Watch, alongside HistoryExtra.

How Immediate decides which brands get a paid product

Importantly, the team hasn’t built a digital product for every brand. Key questions need to be asked before anything is built:  

• Is the audience large enough, and will it return regularly rather than just twice a year?
• Will people habituate to the brand, or is the relationship occasional?
• What can you realistically charge, measured against whoever else already meets that need?

Then the maths: how many subscribers at what acquisition cost, what happens at the first renewal, and what happens at every one after that?

Categories that clear the bar are the ones where Immediate can lead and defend. Food and Wellness through Good Food and Nutracheck, and entertainment through Radio Times. Because each audience answers those questions differently, no two products are priced or packaged the same way.

The same questions explain why paywalling the magazines was never the plan. “If you’re just paywalling content, it doesn’t really take you far enough, unless it’s real must-read content to get people to subscribe and stick around.” Evergreen lifestyle content lacks the urgency of news, so unlocking it alone is not enough. The value has to come from utility instead: cook mode features, prep and planning tools, celebrity takeovers and app-exclusive material.

The websites became the marketing machine

A product that clears the test still has to reach people cheaply enough for the maths to work. Immediate’s answer was already sitting in the business. It kept the free, ad-funded websites that Radio Times, Good Food and the rest already ran, and gave them a different job. They are now the top of a funnel that ends in an app subscription.

“We’ve got big websites for Radio Times, for all of our brands, and we use that to market the products that are appropriate for those audiences,” Burney says. “So it’s a marketing machine, really.” Recipe traffic is directed to the Good Food app, and TV listings to What to Watch. Social reach feeds the same funnel, and paid campaigns run off the back of it.

The machine also runs on creator and influencer talent. Immediate is home to a roster of chefs, presenters and social-first creators, with a dedicated ‘Faces and Voices’ team to cover talent management, including both staff and external names.

“It would be misleading to say that the talent is just going to drive subscribers. The talent is ultimately top of funnel,” Burney says.

Good Food works with talent, including Chetna Makan, Max La Manna, Sophie Wyburd and Angela Hartnett, fronting social campaigns for recipes such as chocolate cake, marry me meatballs and, randomly, rhubarb, at very low cost per order. “I think our best campaign got about 20,000 subs in January.” The same names then run app takeovers that are exclusive to subscribers, so the talent that pulls people in also gives them a reason to stay.

Not all of the talent is bought in. Alex Horne of the UK comedy show Taskmaster featured in an app takeover campaign, and Ricky Gervais did the same for the Radio Times app. So did Perry, a community manager on Burney’s own organic social team, who is massively into musicals. One of his posts about Wicked was retweeted by Ariana Grande and Cynthia Erivo, and passed a million views. 

The talent stays the same down the funnel, but the message changes, from inspiration at the top (cook this dish, here is your Christmas plan, here is what is on TV) to utility at the bottom, the closer someone gets to paying.

One platform, and the commercial discipline underneath it

A funnel that runs from a free recipe page to a paid app subscription needs a technical stack that can manage the entire journey, and Immediate’s stack had grown incrementally. They spent 18 months replacing it. Paywall, billing, revenue platform, and first-party data now sit in a single system driven by a customer data platform (CDP), which was bought and extensively configured rather than built internally. “Building something in-house of this complexity would be out of the means of most publishers,” Burney says.

The result is an AI-driven paywall that the company expects will significantly boost numbers, as well as enable print and digital packaging options which were previously impossible. A separate commercial analysis team holds the budget on what gets funded: acquisition cost against lifetime value, three to one on a digital product over a couple of years. “People either overinvest and see a very poor return and then get disheartened and give up, or see a really good result but then don’t realise that it’s not replicable.”

What good looks like

Subscription only works if you know what a healthy product looks like. “Talk to some subs people, they can get lost in the detail that sits behind it,” Burney says. “But it’s a science and an art at the same time.” A free or low-priced introductory trial would be expected to achieve a 40-55% conversion to paid status. Below that, the return will not be there. The next renewal period should land between 65 and 80%. For monthly retention, anything under 90% is a problem, with 95% as the target.

“Typically, people don’t invest enough in retention. When you’re in growth mode, acquisition is the lifeblood, and that’s also the thing that everyone loves because it’s all high numbers coming in.”

Where the next 1.8 million comes from

The 3 million target rests on international expansion for Good Food and Nutracheck, particularly in the US, as well as organic growth, new products in markets Immediate already knows, and bundling. On bundling, she personally is more sceptical than many other publishers that this has to be one-size fits all. Nutracheck’s growth came from cross-selling into the Good Food audience and applying industrial subscriptions expertise to a small startup.  She notes, “While we are now experimenting with bundling, it’s actually cross-selling.”

“Your success in the current year hasn’t got a lot to do with what you’re doing in the previous year, because the subscribers you acquire are really building future value. What you’ve done in previous years is what’s making today work,” Burney says. “It massively increases the predictability of future earnings. Massively.”

What this means for you

Most of what worked here is separable from Immediate’s scale. The company refused to trade print for digital revenue, so nothing had to be cannibalised. It ran a product test before a content test: would people use the product repeatedly? And was the commercial model tight enough to identify a repeatable win from a one-off?

Publishers should take different things from their experience.

• Single-title magazines: the portfolio logic does not apply, but the product test does. Ask honestly whether your audience would habitually use an app, and whether the utility you could build is worth paying for. If not, an advertising model might be a legitimate answer.
• Regional and local news groups: the estimation discipline and the retention benchmarks travel directly. So does the argument for protecting a profitable print budget rather than forcing migration to a digital solution, and creating something new for audiences
• Large portfolios and media houses: the transferable strategy is refusing a single payment model across brands. Segment by whether the audience is sticky, whether the brand leads its category, and what the build costs, then invest only where all three hold.
• Nationals and news publishers: Burney’s exception is your advantage. News is one of the few categories where paywalled content alone converts, which makes utility a retention question for you as well as an acquisition one.

I’ve run this process with several businesses this year, and it holds up every time. You don’t decide what the subscription includes and test it against the market. You pull apart your own traffic and behavior data, and only decide whether to keep building once that data has told you there’s something worth building. Here’s the process.

1. Start with frequency

The first thing I want to understand is what people already come back for.

For a publisher, that might be a morning briefing, live coverage, a specialist vertical, or a particular columnist. Pull the behavior of your most engaged users and look for repeat patterns. Which products correlate with higher visit frequency? Which newsletters reliably drive people back to site? Which features appear disproportionately among your most active readers?

You’re looking for evidence of habit. A large audience visiting once or twice a month may still support a strong advertising business, but it’s weak evidence for a recurring consumer proposition.

2. Use registration as the next test

Before asking for a credit card, see whether people are willing to identify themselves.

Registration introduces a small amount of friction and asks the user to exchange something of value, usually an email address and some data, for a better experience. That makes it a useful halfway point between anonymous consumption and payment.

Look at registration conversion by audience segment, then compare registered and anonymous users on frequency, depth of engagement and eventual propensity to subscribe. If your most engaged readers still resist creating an account, that’s worth understanding before a payment step goes in front of them.

If registered users who adopt a particular feature come back twice as often, that is a much more useful metric for proposition design than a brainstorm full of ideas about premium benefits.

3. Build paid around behaviors that already exist

Once you know what your most engaged registered users are doing, the paid proposition should give them materially more of it. That might mean greater access, deeper utility, more convenience, exclusivity or a better experience. The important part is that the benefit connects to something the audience has already demonstrated it values.

If live coverage drives unusually high frequency, ask what a paid live product could add. If specialist newsletters show the strongest retention signals, deeper analysis or additional editions may have more value than an unrelated bundle of perks. If users who save articles and follow topics return more frequently, better personalization may be more compelling than another piece of “exclusive content.”

This is also where proposition bloat creeps in. If the pricing page needs six unrelated benefits to make the subscription feel substantial, that’s usually a sign no single benefit is compelling enough. The pricing page should be the output of the proposition work rather than the place where the proposition gets decided.

4. Be deliberate about what remains free

Publishers have an added complication because the free product often drives discovery, advertising revenue and the future subscriber funnel.

Moving too much behind a paywall can reduce the audience available to convert later. Leaving too much outside makes the paid proposition harder to distinguish. The answer depends on what role each piece of content or product plays in the journey.

A story that brings thousands of first-time visitors into the funnel may have a very different job from a specialist product consumed repeatedly by known users. Treating them identically because they both sit on the same website is usually where the trouble starts.

5. Forecast before you keep building

Most subscription models get validated backward. Leadership picks a revenue number it needs, and the team keeps building a forecast toward it without checking whether the underlying audience can actually get there. The proposition work continues on momentum, not evidence.

Build a forecast from the same data steps 1 to 4 produced instead. Take your traffic and frequency segments, apply a realistic registration rate, apply a free-to-paid rate benchmarked against comparable products at a similar price and frequency, then apply a retention curve for that price tier. You’ll get a range, not a number. Compare it to what the business actually needs. If the range falls well short, that’s the signal to stop building and rethink before the pricing conversation even starts. Better to find out before another quarter of work goes into a proposition the audience was never going to support.

Do the work in steps one through five before you launch anything. Some of what comes back will tell you to stop, and that’s a better outcome at step five than a year into a proposition the audience behavior was never going to support.

If those behaviors aren’t there, you’re not a subscription business. Yet.

Jonathan Gupta-Buckley founded The Twenty One and Kaai after a decade leading audience and subscription growth inside News Corp, including building the largest paid subscriber base in Wall Street Journal history. Today, TTO embeds senior operators inside the world’s most ambitious businesses to solve growth challenges in practice, build internal capability, and leave teams with systems they can run themselves. Kaai is TTO’s proprietary AI growth analytics tool. It is currently deployed at a number of global media businesses. 

Why Le Monde invests in YouTube

Lou Grasser reminded the room that Le Monde has been investing in platforms for more than 15 years, with one conviction: offer free content to younger audiences and bring them toward subscription. Le Monde now counts 700,000 subscribers, 90% of them digital. On YouTube specifically, the audience wants ambitious, long formats. In 2025, total views grew 25%, pulled up by long pieces (Charlie Hebdo, the 13 November attacks, the anniversary of Hiroshima). One telling number: followers represent only 25% of total views, so the platform remains a discovery machine rather than a loyalty channel.

“We’re often asked why Le Monde doesn’t invest in a TV channel. At Le Monde, we’d rather invest in the newsroom and in innovative video formats than in something with no direct return on investment.”

Lou Grasser, CDO, Le Monde

Le Monde does 4 million video views a day, against 6 million daily site visits. Much of the video work relies on OSINT (open source intelligence). The long-term goal: make Le Monde a reference for explainers and long formats, while gaining reach through short formats for discoverability.

Pedro Pina: YouTube’s case for news

Pedro Pina rejected the “Death Star” metaphor raised at the open: “YouTube wants this space to flourish, not destroy it.” He shared a few figures and points: 15 billion news views on YouTube last year, with news consumption growing steadily; Gen Z consume a lot of news, contrary to the cliché, and are not limited to short formats or mobile, since the connected TV is a major use. The revenue share leaves the publisher more than half of what it monetises. The moderation noted a recent partnership with France Télévisions for 20,000 hours of news a year on the platform.

On shifting habits, Pina described a move in demand: users already know “the what” through end-of-day notifications. What they want is “the why,” and how it concerns them. Hence, he argued, the explosion of podcasts. He cited the NYT podcast: “you know the news, now we explain why.”

“People expect journalists to have a face, a voice, to be on camera, to have a direct relationship with viewers. That’s not really new in TV. But it’s becoming more and more central.”

Pedro Pina, VP YouTube EMEA

His advice to publishers came in three points: be personality-driven, direct-to-camera and authentic (the hardest lever to move culturally, but the most decisive); don’t copy-paste TV onto YouTube, since it needs a different edit, a different rhythm, different voices; and “hire a YouTuber,” meaning one or two people who have run their own channel and can train the in-house team as it scales. Pina later clarified that he meant channel management, not putting creators in front of the camera in place of journalists.

What Le Monde’s audience actually asks for

Asked live what users want from Le Monde, Lou Grasser listed: long formats, long documentaries and journalists on screen; transparency around investigations, with “behind the scenes” coming up often; short breaking-news formats with systematic voice-over and subtitles; and recognisability of the Le Monde brand, on both long and short content.

On performance, the best-performing video cited was a storytelling piece on the Paris attacks: around 20 minutes of average watch time, more than 2 million views on the English edition and roughly 2.5 million on the French one. Long investigations and explainers perform on reach, engagement and watch time at once. Breaking-news formats, by contrast, are effective for discovery and habit-building, but not enough for long-term retention. Future priorities named by Grasser: strengthen hard-news video (with the live coverage of Ukraine and Iran as a reference), launch video podcasts in thematic series (climate, parenting), and improve video coverage of major events. On the recent municipal elections, Le Monde overperformed across the board and was an ultra leader on search, the weak spot being video and social platforms, hence the priority given to video.

Building one integrated newsroom

Le Monde shifted to a majority of digital subscribers back in 2017, but its workflows were still partly organised by medium. After study trips to American and British newsrooms, Grasser said the newsroom is now fully dedicated to digital, with a small print-specific team. The next step: getting the audio, video and other sections to genuinely work together, as an integrated newsroom rather than format-based silos. On creators, the position is clear: it is not a strategic priority for Le Monde, which would rather invest in its in-house journalists, who in Grasser’s view have the potential to carry stories on screen. For other titles in the group’s portfolio, the calculation could be different.

Monetisation: ads first, video paywall next

Pina reminded the audience that YouTube’s base model is advertising, as the historic model of publishers has always been, and that this is where you start. More than 10 monetisation modes exist on YouTube, and subscription products already exist. What is in the works is the integration between YouTube’s paywall and publishers’ own paywall, with partners like Le Monde pushing for it. The open questions: confidentiality, data handling and legal compliance.

“On a long investigation titled ‘Northern Europe is preparing for war with Russia,’ put behind the paywall on the site, we got 300 to 400 conversions in a few days. We can’t paywall on YouTube yet, so we had to publish the video on the platform a few weeks later. It’s a strong opportunity for us.”

Lou Grasser, CDO, Le Monde

12 months and 5 years out

Grasser’s 12-month horizon: deliver first-rate video coverage of the 2027 French presidential election, and make progress on dubbing and translation, notably to boost the growth of Le Monde in English. Her 5-year horizon: a newsroom no longer organised by sections (audio, video, text) but fully integrated, serving readers the content they want in the format they want. Pina’s 5-year view: that every major traditional journalism brand becomes video-first, not because it is good for YouTube, but because it is good for viewers and for society.

Methodology: written for Audiencers by Claude (Anthropic) from a Trint audio transcript of the session, translated and cleaned. Quotations in quote marks are faithful reformulations, not exact verbatims. WAN-IFRA World News Media Congress, Marseille, 3 June 2026, Partner Showcase.

We have entered the Age of Experience

Audiences still seek information and fulfill particular needs. But they increasingly choose where to spend their time according to the experience surrounding that value and whether another worthwhile experience follows.

Where platforms provide a compelling sequence, attention accumulates.

Experience is more than a new way to describe “what users need.” It is a different way to organize what an outlet produces, how its products connect, and how audience value translates into revenue.

Asking “How can we improve this product?” therefore misses the sharper question: “What experience should audiences consistently associate with us, and what should lead them to the next one?”

Experience connects audience strategy with revenue

Publishers often move quickly from an audience goal to an output.

“We need to reach younger audiences” becomes “we need more vertical video.” “We need stronger habits” becomes “we need a daily newsletter.”

Those may be sensible decisions, but format alone does not explain why audiences will value the product, return, or move deeper into the publisher’s environment.

This is the logic behind the AXR framework: Audience → Experience → Revenue.

We developed AXR to connect audience strategy to revenue through the experience layer. That layer defines the audience moment, the promise, how the product fits into a routine, and the behavior it should create. Revenue can then follow through advertising, sponsorship, registration, or subscription.

For example, “a daily video briefing for younger audiences” is still mainly a format description.

A stronger definition would describe the audience moment, why the briefing is worth returning to, what should follow it, and which elements, such as the host, pace, promise, or format, must remain recognizable across the sequence, beyond this particular item.

That difference matters, because formats produce outputs. Experiences create behaviors. Repeatable behaviors (staying longer, returning more often, and moving to the next item) connect audience value to commercial value.

The most valuable experience is the next one

One strong experience is rarely enough.

Audiences do not open a social platform for one excellent post and then move elsewhere. They stay because another relevant experience is immediately available. Games create the same continuity from one level or challenge to the next.

Publishers still tend to optimize the individual item. Even when related content is offered, the connection is often topical rather than experiential. The next item may cover the same subject but feel completely different in format, pace or purpose.

In our work developing recurring video experiences, we see a passive-active pattern. After one video, many viewers prefer to remain passive and let the next one begin. But when the experience changes too sharply, they must become active again: reassess, choose, and decide whether to continue.

That decision point is often where they leave.

Continuity does not mean showing people more of the same. It means preserving enough of the experience promise while introducing new value.

Before launching a new product or format, publishers should therefore ask:

• What experience begins here?
• What is the most natural next experience?
• What must remain consistent so the next experience feels like a continuation rather than a new decision?

These questions change the unit of planning. Instead of producing one engaging item and hoping the audience finds another, the publisher begins designing the sequence itself.

The bottom line

The strategic question is no longer only which products to build, but how those products combine into a recognizable experience system.

The practical shift is to evaluate every product on two levels: the experience it creates on its own, and the role it plays in the sequence around it.

Publishers that design both levels can turn journalism into more time spent, stronger habits, and clearer paths to revenue. In the Age of Experience, the advantage will belong to outlets that make every experience valuable, recognizable and naturally connected to the next.

Inside, you’ll find data on how engagement concentrates among a small group of super users, how much richer content formats like audio, video and games boost engagement and where publishers are still struggling to measure the value their apps create.

Here are a handful of the biggest takeaways from this year’s report, useful for any publisher trying to get a deeper understanding of how their apps are performing and where to focus effort going forward.

> The Pugpig Media App Report, drawing on 440 apps and 140 publishers, finds engagement is wildly concentrated. 15% of monthly users generate 75% of all sessions, and 6% account for 75% of screen views

> Games, audio and video are the engagement multipliers. Players open the app 2.7 times more often, video watchers spend 3.3 times longer in it, and audio listeners 2.5 times longer

> Between 11% and 94% of a news publisher's subscribers open its app in a month. Yet only 26% of publishers say the app's main job is acquiring subscribers

> Publishers agree on what matters: retention, acquisition, time spent. Far fewer can measure it consistently, which makes it easy for the app's contribution to be undervalued internally

> The biggest blocker is not budget or technology. Half of publishers name product team capacity, and the strongest apps are simply the ones given sustained attention rather than clever engineering

Rich features consistently drive deeper engagement

Features like games, audio and video consistently drive deeper engagement. Users who interact with these formats, as well as those who search, save and share articles, show a clear lift in engagement compared with those who do not.

Users who play games open their app 2.7x more often than those who don’t and rack up over 29 hours of in-app time a month in top-performing apps. Video and audio show a similar pattern, with users who watch video spending 3.3x more time in-app each month and audio listeners spend 2.5x longer.

Search, save and share tell a similar story, suggesting these are genuine signals of loyalty and intent and important features for publishers to use in their apps. The report shows that these features have an outsized impact on how habitual and valuable that group becomes.

A small number of super users drive the majority of engagement

One of the clearest patterns in this year’s data is how important super users are to overall app engagement. Our report found that just 15% of monthly active users generate 75% of all sessions, whilst 6% of MAUs account for 75% of all screen views. Those numbers point to a small group of users that come back to the app more often and get through far more content once they’re there than the average reader.

This should frame how publishers think about app strategy. Growing the total user base is very important, but the data suggests that significant value comes from deepening the relationship with the most engaged readers. That means identifying the content and features that most encourage repeat use.

There’s a wide gap in how many subscribers actually use the app

Another significant factor that drives app engagement is the share of the audience with a direct subscription, which varies widely between publishers. The report finds that, for news titles, anywhere from 11% to 94% of total subscribers open the app each month. 

There’s a real opportunity here too. Many apps attract a substantial number of non-subscribers alongside their paying base, yet only 26% of survey respondents said the main role of the app is subscriber acquisition. With non-subscribed users significantly less engaged than paying users, there is clearly a lever publishers could pull to improve conversion.

The report also hints at a commercial opportunity in in-app advertising. Apps are often designed as ad-light experiences, which keeps impressions per screen low when compared to web. But high session frequency among engaged users means total impressions per user per month can still be meaningful, particularly when formats and targeting are tuned to app audiences. Treating the app as a revenue platform across both reader and advertising income is one of the clearest commercial signals in the data.

Publishers know what to measure, but not all can measure it well

When we speak to our customers about what matters most for their apps we get fairly consistent answers around subscriber retention, acquisition, time spent and engagement. However, according to this year’s survey, there is a significant challenge in the ability to measure these KPIs with any consistency.

That gap between what publishers know they should track and what they can actually measure is significant. Without clean measurement, it becomes much harder to attribute revenue to the app clearly, which in turn makes it easier for the app’s contribution to be undervalued internally.

According to the report, capacity, not money or technology, is the issue. Product team capacity is named by half of publishers as the single biggest blocker to progress, well ahead of budget constraints, stakeholder alignment or editorial capacity. 

Effort and investment separate the top performers from the rest

Perhaps the most encouraging takeaway from this year’s report is that strong app performance is less about technical sophistication and more about sustained investment. The best-performing apps are usually the ones publishers have consistently given time, resources and attention to.

That shows up across almost every metric in the benchmark data. Richer content formats, push strategy, onboarding, personalisation and content discovery all follow the same pattern. The report’s overall message is that small, sustained improvements compound, and publishers who treat their app as a genuine priority tend to see it reflected in the metrics

Next steps for publishers 

For teams looking to act on the data, the report closes with a set of practical recommendations for turning these findings into progress.

Getting more subscribers to actually use the app is the starting point, alongside investing in features that reward repeat use. Audio, video, games, search, save and sharing are clearly all associated with stronger engagement, and there’s clear room for more publishers to make better use of them.

The report also flags push notifications as an underused tool and highlights the opportunity in converting anonymous users into known readers, given how many app users still aren’t logged in or registered.

Beyond acquisition and adoption, the guidance turns to consistency and discovery. Publishing frequently, improving how users find relevant content and measuring app performance against real business outcomes are all highlighted as ways to make the app pull its weight as a genuine engagement, retention and revenue channel.

These findings show that the apps performing the strongest currently are combining richer content, sharper measurement and consistent investment to build real habits with their readers. 

For any publisher trying to work out where they stand, that’s exactly what this year’s report is designed to help with, giving you a clear picture of how your app compares and what the leading publishers are doing differently.

The 2026 Pugpig Media App Report is available to download now, published by Pugpig, the mobile app platform that powers apps for over 450 of the world’s leading media brands.

The premium experience offered in the app

The app isn’t a stripped-down version of the site. It’s built around giving already-engaged readers more ways to spend time with SFGATE’s journalism and journalists.

• Pop-up newsrooms (launching soon) in parks across San Francisco, where readers can meet a journalist in person. Editor-in-chief Grant Marek cited a statistic that only 23% of people have ever met a journalist face to face, and framed the initiative as a direct attempt to change that for SFGATE’s own readers
• Planned live journalism events across California, built around specific local reporters who are, in some cases, literally living the stories they cover. Julie Brown Davis, a reporter covering the Lake Tahoe housing market, was born and raised there, and was herself pushed out to Reno by rising housing costs. The plan is to put reporters like her in community spaces with readers who already trust them
• A California-themed mini crossword, distinct from a generic puzzle in that some clues specifically reward local knowledge. The app adds a streaks feature (consecutive days completed) and a leaderboard letting readers see how they stack up against SFGATE’s own staff
• Audio versions of articles, extensive bookmarking, and premium maps that plot SFGATE’s historic archive geographically
• Granular push notification controls, roughly 30 different alert categories, built specifically because California’s geography doesn’t map neatly onto any one set of interests

Why Friends?

Before landing on ‘Friends of SFGATE,’ the team tested more classic subscription names, including SFGATE+ and SFGATE All Access. But they found that ‘subscription’ read as too transactional for what they were trying to signal, and ‘membership’ carried strong associations with charities and causes that people support out of obligation rather than affinity. ‘Friends’ tested as warmer than the former and less formal than the latter, reflecting a relationship that already existed with SFGATE’s most engaged readers, rather than inventing a new transactional one.

Redefining what success looks like

Marek described a shift in how his newsroom is measuring its own performance since the move towards reader revenue. Total visits, historically the default newsroom metric, matters less to this project than repeat visits. How many times can someone who’s already shown up once be brought back a second and third time. 

That reframing also changes the editorial calculus. While competing on the exact same breaking story as ten other outlets used to be table stakes, Marek’s team is now weighing differentiated, SFGATE-specific coverage more heavily, on the logic that unique journalism is what actually builds the habit loop the app is designed around.

The early numbers

Four weeks post-launch, one figure stood out to Campbell specifically: of the readers who took the 7-day free trial, close to 90% converted into paying Friends. It’s an early signal rather than a settled result, but this number represents a higher trial-to-paid conversion rate than he’s seen in close to three decades of doing this work.

This is, in part, due to active retention early on, during these 7 days:

• An email series that includes a personal message from Marek himself, alongside reminders of specific app benefits
• Push notifications, tuned to each reader’s chosen interests from day one
• The crossword’s streak mechanic, which Marek described as quietly addictive once someone gets a few days in
• Daily social content (SFGATE has the largest Instagram following of any Hearst Newspapers title) introducing different app features on rotation, audio, bookmarking, events, to readers whether or not they’ve installed the app yet

What changed internally to make this possible

The launch required real organizational change, not just a new product wrapper:

• A newly created role, Community Engagement Program Manager, currently being hired for, is a first for SFGATE
• A second hire added to the video team, driven by strong reader response to journalists as recognizable, followable creators (Melissa Cho, known to readers as “SFGATE Girl,” for instance)
• An existing audience desk editor, with prior app programming experience at The Athletic and SmartNews, moved into leading app programming for SFGATE specifically, work Marek described as sitting dormant until this launch gave it a purpose
• On the marketing side, Hearst’s existing team and infrastructure stayed the same in headcount. The real work was calibration, such as learning to write copy and design interstitials that actually sound like SFGATE, a two-way collaboration between the marketing lead and SFGATE’s news team

Marek connected this to a broader shift he’s watched happen over his career. Journalists who started out purely writing have had to add video, then podcasting, and now, in his framing, business literacy generally. Understanding Campbell’s side of the operation, and vice versa, is what he considers necessary now for finding initiatives that contribute to reader revenue. 

What’s still being figured out

On acquisition, Campbell’s team is four tests into learning what actually drives trial sign-ups, testing different feature framings and on-site placements, with paid acquisition still to come.

On audience targeting, Campbell said the current approach has been intentionally broad – go wide and go loud, but the next phase, drawing on lessons from Hearst’s other subscription brands, is sharper segmentation, including identifying which readers are statistically most likely to convert to a Friend, and targeting acquisition efforts accordingly rather than treating the whole audience as one pool.

The bigger shift this represents

Campbell framed Friends of SFGATE as evidence that the old binary (a site that’s either a subscription business or a free one) doesn’t have to hold. 

It’s a structurally similar idea to the ‘willingness to pay versus willingness to commit’ gap The Washington Post has written about in its own flexible-access research. Rather than forcing every reader through the same single gate, SFGATE has effectively added a new, optional tier above the free web experience, built specifically for readers who were already deeply engaged but never had a way to formalize that relationship or directly support the journalism.

SFGATE has, for its entire history, had no answer for the reader who finishes a story during a big moment and asks how they can support their work. Now there’s a concrete answer. And one that has been designed with lasting reader relationships in mind. 

Thank you to Grant & Mark for their time!

That leaves a whole category of question permanently unanswered. The recurring ones, where we need to know:

• How a piece of content will serve the users we are trying to reach
• Which of several options is weakest, before we invest in any of them
• What objections a change is likely to run into
• How different groups of users will react to the same thing

To take one example: whether a new onboarding flow feels like a data grab. Too small to commission a study for, and it creates uncertainty that festers within the onboarding team. So somebody decides on instinct, usually whoever is most senior, and the instinct becomes the decision. It might well be right. Nobody will ever know though because nothing was tested.

This is the gap synthetic audiences fill. They don’t have to replace the research we already do well. What they change is the dozens of small calls a month that currently run on somebody’s hunch. And for the many newsrooms doing no audience research at all, the comparison isn’t between synthetic feedback and properly-designed research. It’s between synthetic feedback and nothing, and synthetic feedback done well beats nothing.

The value is that a whole class of decisions moves from opinion to evidence, at a cost and speed that makes asking routine.

What a synthetic audience is

“Synthetic audience” is an umbrella term, and the field hasn’t settled on what sits under it. You’ll see people use synthetic personas, synthetic users, digital twins, synthetic panels and silicon samples loosely and sometimes interchangeably. I’m using three here because the differences between them change what we can trust.

• A synthetic user is an LLM given a description of a type of person and asked to respond as they would. We write the description – the segment’s age range and media habits, whatever we know – put it in the prompt, and ask the question. Nothing sits underneath it. What comes back is the model’s averaged sense of what somebody matching that description would say, assembled from everything in its training data about people like that.
• A digital twin is built from one real person’s own material: an interview transcript, their survey answers, their support emails, their behaviour on our site. That material goes into the prompt and the model answers new questions as that person. It’s also possible to fine-tune a commercially available LLM on the material instead of putting it in the prompt. Academic studies from Stanford and Columbia tested that route, and neither found it beat simply placing the person’s words in the prompt. Which is good news for small teams: building a twin should not require any substantial engineering lift.
• A synthetic panel is either of the above, multiplied. We can build twenty or fifty across our segments, put the same question to all of them, and read the pattern rather than any single answer.

What synthetic audiences are good for

A number of use cases have emerged. These four have the clearest evidence behind them:

Testing language before it goes out. Paywall messages, membership appeals, onboarding copy, campaign framing. We put the draft in front of a synthetic panel and ask what they find unclear and what would make them stop reading. This is the highest-value use, partly because language is what these models handle best, and partly because it’s a decision we make most often.

Narrowing a field of options. Six newsletter concepts, or four pricing structures. The panel won’t tell us which one to launch. But it will reliably tell us which ones are weakest. So, the real research budget goes to the survivors. Spindrift Beverage Co., a US drinks brand, used a synthetic panel to choose between new product concepts. It reached the same answer in a week that a 500-person consumer study had taken two months to produce.

Surfacing objections we haven’t thought of. Ask the panel what would stop them subscribing, or what they’d assume about us from our about page, and we get a list of concerns to check. Where synthetic panels have been compared against real research, they have identified the same themes and the same objections. The magnitude and the proportions differ, the list doesn’t.

Comparing groups against the same thing. Put one paywall message to a synthetic lapsed-subscriber panel and a synthetic loyal-reader panel, then read the difference. Relative comparison is where these systems are most useful.

You may have spotted that all of the four use cases narrow or pressure-test rather than decide.

How accurate are synthetic audiences

A figure of around 85% circulates in coverage of this field, usually without much explanation. Read quickly, it suggests a synthetic version of a person gets 85 out of every 100 answers right. That isn’t what was measured.

The figure comes from the Stanford study. Researchers interviewed just over a thousand people at length. They built a digital twin of each person from their interview transcript. Then they tested whether the twin could predict that person’s answers to a survey the person had already completed.

But they didn’t score the twin’s answers against the persons’ answers and leave it at that. Because if we ask a person to fill the same survey again two weeks later, they match their own earlier answers only about eight times in ten. (Memory shifts!) So, the fair question is not whether a twin matches perfectly. It’s how a twin does against that eight-in-ten human benchmark.

Researchers found that the synthetic twins reached roughly 85% of that human benchmark – so, not 85% of the answers.

The figure tells us that synthetic twins built on rich individual material carry genuine signals about the people they came from.

The evidence in this field is still early, and most of it comes from outside media. For me, that’s another reason to use synthetic audiences for narrowing rather than deciding, which is what the rest of this piece argues for.

Where synthetic audiences stop

They can’t tell us how much. Synthetic responses cluster. Real people spread out more across opinions, while synthetic ones converge towards each other. A panel can tell us which option comes out ahead. It can’t tell us by how much, which rules out adoption rates, willingness to pay, satisfaction scores and anything else where the size of the number is the finding.

They can’t show us the small segments that matter. Clustering also means that the users at the edges disappear: the share of subscribers already on their way to cancelling, or the small core of superfans who account for a disproportionate slice of revenue. Those groups are small because they’re specific, and specificity is something a synthetic panel loses.

They tilt in a consistent direction. Synthetic respondents tend to be WEIRD (Western, Educated, Industrialised, Rich, and Democratic). They come out more comfortable with technology, less worried about privacy, better informed and more agreeable than the people they’re modelled on.

They shouldn’t speak for communities we haven’t spoken to. A team in the US building a WhatsApp media-literacy service for the Indian diaspora tried synthetic personas and got the diaspora as Westernised media imagines it. They caught it because they knew what the actual community sounds like. This is the kind of stereotype that synthetic responses can perpetuate.

How to start experimenting with synthetic audiences

Running a synthetic panel takes four steps, and each exists to make the next one worth doing.

The first fixes what you’re actually trying to settle, so you don’t generate material nobody acts on. The second decides what evidence the panel stands on, which sets how much weight its answers can carry. The third picks the kind of question you’re putting to it, which determines whether the answer is usable at all. The fourth gets an honest answer out, then records it in a form a colleague can judge for themselves rather than take on trust.

Set up once, the marginal cost of the next question is close to nothing. That’s the point. You’re building something you put questions to whenever they come up.

1. Fix the decision before you write the question

Most synthetic research fails before a panel is ever built, by producing interesting material nobody was going to act on in the first place.

Understand what you’re trying to achieve – what decision you’re hoping to make. Then determine the insights you need to inform the decision. Then figure out how to get those insights from a synthetic panel.

This ensures that the work you’re going to do is aligned with what you’re working towards. This is the step that stops the output becoming a slide nobody uses.

2. Decide what data the panel is standing on

What a panel stands on sets the ceiling on everything after it. The same question put to a well-grounded panel and an ungrounded one produces answers that may look similar but instead mean completely different things.

There are three levels:

• Nothing. A prompt describing a reader, with no evidence underneath it. What comes back is the model’s general impression of people like that.
• Demographics. Market or audience data about the segment you’re aiming at. Better, and still a description of a category rather than of anybody.
• Real material from real readers. Words that actual people you can name said or wrote.

The test is whether you can name the humans whose words are in the system. If you can’t, you’re on one of the first two levels.

Reaching the third level may be easier than it looks, depending on what your newsroom has kept. Worth looking for:

• Past interview transcripts, from any research project, however old
• Open-text answers from surveys (which usually go unread)
• Subscriber cancellation reasons
• Support and complaints emails
• Comments and community posts

Prepare the material by anonymising the person and keeping the detail that makes them specific: their circumstances, and what they came to you for. Spread it across your segments rather than piling it up. Fifteen people covering five reader types beats forty of the same type.

3. Pick the job

There’s more than one kind of question you can put to a synthetic panel. Two are worth setting out here, because most teams use the first but never the second. Choosing deliberately between them is what separates a usable answer from a merely plausible one.

• Count them. Put the same question to the whole panel and use the aggregate: which of six concepts is weakest, or which objection comes up most often. Read the ranking and ignore the margins. This is the workhorse, and it’s subject to the tilt described earlier.
• Analyse against them. Instead of asking how users would react to a membership appeal, ask where that appeal conflicts with what you’ve documented about them. The model isn’t being asked to be anyone. It compares your draft against a profile you wrote and can check, then hands back specific points of friction you can verify or throw out. Most teams never try this, and it’s the job best matched to what these systems genuinely do well, which is analyse language.

4. Make it argue back, then record what you did

Left alone, these models are agreeable, and an agreeable panel tends to tell you what you already believe.

Three things to build into the prompt:

• Tell it to answer as a reader rather than as an analyst. Otherwise the responses arrive sounding like a dull report, with tidy conclusions no human user would share.
• Tell it that being indifferent is a valid answer. Plenty of readers genuinely don’t care, and a panel where everyone holds a strong view is already wrong.
• Show it a few examples of how real people answered a similar question, with any polished phrasing taken out.

Then narrow with the panel and validate the surviving one or two options with real people. You’re only testing what has already survived something, which is where the saving comes from.

Finally, write up what you ran: who the panel was defined as, what it was grounded in, what you showed it, what you asked, what came back, and what still needs checking. Synthetic output is convincing whether or not it happens to be right. In one study, experienced researchers reviewing persona material could not reliably say which of it had been written by people. Making the output less convincing doesn’t help either, because results that read as obviously machine-made don’t get acted on. Since you can’t remove the believability, the write-up is what stops a plausible answer becoming a fact by repetition.

Start with the research you already have

The case for synthetic audiences isn’t that they closely match real readers. Academic testing has consistently found that they don’t. The case is that most audience questions in most newsrooms are currently answered by whoever is most senior in the room, and a synthetic panel grounded in things our actual readers said is better than that.

So, use them to narrow, not to decide. Ground them in real material, and know which level you’re on. Ask them to compare and to analyse. Keep them away from magnitudes, from the small segments that may matter most, and from communities you haven’t met. And, finally, validate what survives with people.

To get started: find the research your newsroom already commissioned and hasn’t opened since – the launch interviews, the subscriber survey, last year’s churn study – and work out which of this month’s decisions it could answer, if only you could ask it questions.