BBC Studios had a genuinely hard problem: grow a brand-new subscription business in the US, its first ever outside the UK, without damaging the free, ad-supported reach that funds the rest of the business. Block too much content too fast, and you risk both your advertisers and the audience trust BBC has spent a century building. Move too cautiously, and the new business never gets off the ground. 

"We really had to learn to walk first before we could run. In fact, we needed to learn to crawl," – Stephen Macleod, VP Analytics and Measurement, who led the data work behind BBC's new subscription. 

That's exactly what BBC Studios' team did: crawl, then walk, then run, testing every step before committing to it. In June 2025, that work became bbc.com's first-ever paid subscription in the US. Here's how they got there, why it worked, and how you could do something similar. 

What they did 

• They tested the smallest possible change first, before touching anything else. Instead of building a whole paywall on day one, they simply added a prompt asking readers to register, with no new content and nothing locked. "The initial results exceeded our expectations, going from practically nothing to 100% uplift on day one," Stephen said.

  They then went a step further and find out who was actually likely to subscribe before asking anyone to pay. Behind the scenes, they trained a model on reader behavior that could estimate how likely someone was to subscribe, without changing anything the reader could see. That alone doubled daily registrations again, giving the team real, tested data to bring to a business that was, understandably, nervous about the idea. It worked so well, in fact, that they deliberately paused the experiment: a real paywall was coming in a couple of months, and they didn't want to ask the same readers to do something else so soon after.

• Data became unified. BBC had already used Piano Analytics for more than five years, so when they added subscription data, it flowed into a single warehouse alongside everything else, including their own login data from BBC Account. One team, and increasingly the wider business, could now work from the same numbers instead of separate, conflicting reports. 

• Before rolling it out widely, they tested every version of the paywall. One test showed how many free pages a reader could see before being asked to subscribe, at different strictness levels. The surprising result: highly engaged readers converted just as well whether they saw three free pages or seven.That let the team tighten the wall for these readers without losing subscribers. It's also exactly the kind of trade-off they watched closely every time, since showing fewer free pages to a reader also means fewer opportunities to serve them ads before they convert.

• They started using the data to target readers more precisely. Reader segments now flow into a customer data platform that syncs with Salesforce and Meta, helping the team run more targeted marketing campaigns instead of one-size-fits-all outreach. A content recommendation pilot is next: using what someone reads or watches to suggest what they'd want next, in real time. They're now working on automating the hardest trade-off of all. Right now, deciding whether to show a reader another ad or ask them to subscribe is still a manual, team-by-team judgment call. The next step is using AI to make that call automatically and in real time, for every reader, every time. A separate project is building a single visual map of how every metric in the business, from content published to ad impressions, ladders up to overall revenue, so a change anywhere can be traced all the way to the bottom line. 

Bottom line

• Launch an MVP first. BBC didn't start with a paywall. It started with a two-page prompt and no content changes at all, and that alone taught them more than a bigger, riskier launch would have this early. "One of our guiding principles was to test everything," Stephen said. 

• Understand your audience before you ask them to pay. The propensity model ran invisibly in the background, giving the team real signal on who was likely to convert before a single reader was asked to pay. 

• Don’t optimize for just one metric. Every test was checked against its impact on ad revenue, not just subscriptions. 

• Create one shared source of truth. When ad data, finance spreadsheets, and warehouse data all disagree, the fix usually isn't a longer meeting. It's building toward one shared source of truth everyone pulls from. 

Where Piano fit in, and how you can start 

Piano has actually been part of this story from the start. BBC has run Piano Analytics for more than five years, which is exactly why adding subscription data into one place was, in Stephen's words, "a kind of seamless transition." The warehouse holding all of it runs on Snowflake, built and managed as part of BBC's Piano setup. Composer powered the registration prompts and paywall experiments described above, from the very first two-page test to the live paywall. 

If you're earlier in this journey than BBC, the good news is you don't need all of it on day one. You can start with real-time analytics on your site or app, add a simple registration or paywall test whenever you're ready, and grow into a shared data warehouse and audience targeting later, the same crawl-walk-run order BBC followed. And if you're further along and already juggling ad revenue against subscriptions by hand, that's what Piano's Ad Revenue Insights is built to automate: using AI to weigh that trade-off in real time, for each reader, instead of leaving it to a person's best guess. 

This talk was originally given at Piano Academy 2025. If you want to hear more presentations like it, on AI, data, and revenue growth from teams putting it into practice, join us at Piano Academy 2026 in Amsterdam, October 6–7. Register to save your spot.