Run paywall experiments for long enough and you eventually hit a wall: the winning design just keeps winning, and every new variation lands within the margin of noise. That's where Nick Godwin, our head of customer growth and product, found himself at the start of this year. At Superwall, he runs around 200 experiments annually across roughly 60 large and hyper-growth apps, and the design-your-trial paywall had been so durable across app categories that improving on it felt like spinning in place.
In our most recent office hours session, Nick walked through the thesis he landed on after hitting that wall, and the three bets he thinks separate apps that grow in 2026 from apps that plateau.
Bet 1: Move analytics from dashboards to AI agents
The first bet is freeform analytics, shorthand for using an AI agent to ask the kind of questions a dashboard was never built to answer. With the quickly shifting landscape of AI, a fantastic use-case has emerged. Without question, it's deploying these LLMs on as much of your app's data as you can.
Why? Beacuse answering in-depth questions, across several different data domains, becomes as easy as a prompt. Questions that took a fair bit of digging are accessible now, such as:
Are my special-offer paywalls cannibalizing revenue or are they actually additive?
Which price variant is the net ARPU winner across multiple touchpoints?
Which combinations of onboarding inputs correlate with the highest conversion rate, ARPU, and retention?
On a normal dashboard you need to know your own event taxonomy, find the exact chart, pull the right filters, and QA the output. With an agent, you can ask in plain language and get a defensible answer in minutes.
When we polled attendees on what they use for deep monetization analysis, Amplitude led at 41%, with Mixpanel and PostHog tied at 24%. Zero people picked Claude, Codex, or another agent. Superwall.ai is built for exactly this kind of analysis — it shows its work, hands you the underlying CSV so you can vet the rows yourself, and lets you refine the prompt in conversation when it misreads your intent on the first pass.
Bet 2: Personalize the whole funnel, not just the paywall
Personalization usually gets watered down to mean "tweak a paywall headline." Nick's version, as always, goes a bit further. To Nick, personalization means routing users through fundamentally different onboarding and paywall flows. Intent and behavior should drive the segmentation; demographics, in turn, support it where they help.
A few examples from the session:
An AI phone assistant app discovered that work status was their single biggest differentiator. Business owners and self-employed users had an 85% higher conversion rate and accounted for 75% of realized revenue, which meant the team could put its experimentation effort exactly where the money was.
A restaurant menu guide found that users who eat out three-plus times a week converted so quickly that a long design-your-trial offer underperformed a tighter three-day trial at higher pricing for that segment.
A gym app ranked acquisition source, user goal, and experience level as their top three RPU drivers, then built distinct onboarding flows and UGC hooks around the resulting archetypes.
And finally, a social app that opened with 12 onboarding questions found only three actually correlated with conversion. Cutting the rest lifted onboarding completion materially for low-friction segments, while the longer flow stayed in place for users who genuinely needed it.
Bet 3: Get serious about app-to-web (and don't sleep on web-to-app either)
Web to app and app to web are not silver bullets, but they can be a valuable tool in your monetization efforts. So, bet three was centered around web payments.
Nick has run app-to-web experiments with more than 50 customers, and his pushback on how most teams approach it bears repeating. App-to-web is not a single all-or-nothing lever. Wiring up a Stripe checkout and hoping it lifts revenue across the entire user base is the wrong move. Use freeform analytics to find the segments where web checkout actually fits: qualified users where Stripe friction won't stop them, churned users you're retargeting via email, and organic-search traffic that arrived as comparison shoppers rather than through an ad.
If you dive into the numbers a bit, I think most would agree that the economics are what make the bet worth running. Initial conversion versus in-app purchase can drop by as much as 37% because of the added Stripe friction, but trial-to-paid lift usually runs 50% or higher, and average ARPU lifts land in the 11% to 27% range.
One (important!) caveat: if you're on Apple's Small Business Program paying 15%, app-to-web typically nets out roughly flat, so stay where you are until you graduate to the 30% tier. Web-to-app is a longer build, but the cannibalization fear that holds most teams back is mostly unfounded. The overlap between users who would have converted in-app versus through a web funnel is only 5% to 7%, meaning these are different users you weren't reaching anyway. Pick a formula that's already working for someone, start collecting data, and worry about segmenting your funnels once you have the volume to do it well.
Where to take this next
Segment first, optimize second is the thread running through all three bets. Freeform analytics is how you find the segments. Personalization is how you treat them differently once you have them. App-to-web is one of the higher-margin levers you can pull for the segments that respond to it. Any one of the three should move your numbers, but if you stack them together then the compounding starts.
Catch the full session
The design-your-trial paywall template Nick referenced lives in the high-converting templates group in the Superwall library if you want to test it before anything else. And if you have questions about wiring any of these bets up in Superwall, our team is around — drop into our next office hours and bring them.
Want the slides from the session? Grab them here.