Opal's Kenneth Schlenker on focus: one metric, an early paywall, and an iOS-first launch
Opal is a screen time management app that, according to founder and CEO Kenneth Schlenker, reached $5M in annual recurring revenue with a team of seven. The path he describes is deliberately narrow: pick one number to steer by, charge from the start, ship on one platform, and test constantly — but not endlessly. This guide restates those choices as decisions you can evaluate for your own subscription app, with a checklist and a worked example rather than a transcript summary.
Why a narrow playbook can beat a broad one
App teams face far more variables than they can act on at once: acquisition channels, pricing, platforms, onboarding, retention. Spreading effort across all of them at the same time makes it hard to tell what moved the result. Opal's reported approach was to constrain the variables on purpose — one metric, one platform, one monetization stance — so that each iteration had a clear signal. That is a discipline choice, not a claim that every app should copy it. The sections below separate the four decisions and note when each one stops being the right default.
Steering by a single metric: day-8 ROAS
Opal spent on paid acquisition early and treated 8-day ROAS as the indicator of whether growth was working. The logic is simple: if a cohort of new users does not return enough revenue within eight days to cover what it cost to acquire them, scaling spend just multiplies the gap.
A worked example using that rule: suppose a campaign costs $10,000 and the cohort it brings in produces $7,500 in tracked revenue by day eight. Day-8 ROAS is 0.75 — below break-even on that window — so the decision is to hold or fix the campaign rather than increase budget. If the same cohort returns $12,000, day-8 ROAS is 1.2 and scaling becomes a defensible next step. The number itself is not a universal target; the value is that one agreed threshold turns ad-spend debates into a repeatable rule.
Choosing one metric does not mean ignoring everything else. It means one number owns the go/no-go decision while other data stays diagnostic.
Charging early to test whether people will pay
Many apps grow a free audience first and add subscriptions later; Duolingo is the example raised. Opal did the reverse, launching with a relatively expensive annual subscription and later moving toward freemium to reach new groups such as college students. Schlenker framed the early paywall as a discovery engine: payment is evidence that the product solves a problem people value, and it validates product-market fit before you invest in a large free base.
The two models answer different questions. An early paywall asks "will anyone pay for this now?" A freemium build asks "how large an audience can this reach, and what share will convert?" Neither is wrong, but the first produces a faster, harsher signal, and the second requires more runway before it tells you anything. If you cannot yet afford a long free-growth phase, the early-paywall route gives you a verdict sooner.
Launching on one platform first
Shipping on iOS, Android, and web at once multiplies engineering and design work without reliably doubling revenue. Schlenker's argument is that a single platform keeps the discovery phase simple: with two platforms live, you are committed to both and lose the freedom to iterate quickly on one. He cites founders who named an early Android or desktop launch among their biggest regrets.
Use this as a conditional rule, not a permanent one. A single-platform launch fits while you are still searching for fit and your target users concentrate on that platform. It stops fitting when the missing platform is where your paying audience actually is, or when a competitor is locking in that audience first. The long-term plan can still include every platform; the question is sequencing, not exclusion.
What 121 A/B tests taught the team
Opal ran a large number of experiments, and the findings below are the ones the team highlighted. Treat them as hypotheses worth testing in your own funnel, not settled rules — results depend on audience, price point, and product category.
- Put the paywall in front of users early rather than deep in the flow.
- Onboard new users into a scheduled session instead of leaving the first action open-ended.
- Use tiered referral rewards so heavier referrers gain more.
- Recognize that some paywalls, such as the Blinkist layout, are hard to beat; if a proven design is working, changing it needs a reason.
- Limit the number of choices in onboarding; too many options slow decisions.
- For re-engagement, a discount is usually the strongest lever.
Schlenker's own closing advice runs against the headline number: you probably do not need 121 tests. Fewer, better-targeted experiments with clear hypotheses will usually teach you more per unit of effort than a long list of small variations.
A checklist before you copy any of it
Run through these before adopting the playbook:
- Can you name one metric that owns your go/no-go decision, and is its measurement window short enough to act on?
- Does your price point let an early paywall produce a meaningful signal, or will too few users pay to learn anything?
- Is your target audience concentrated on one platform today?
- Do you have a written hypothesis for each test, and a stopping rule for when you have enough tests?
- Have you defined what would make you add a second platform or switch to freemium?
If several answers are "no," the narrow playbook is premature; fix the measurement and audience questions first.
Where to go deeper
The full conversation is available as audio and video, and the related reading below covers paywall placement, measurement pitfalls, and retention messaging in more detail.
| Choice | What it does | Fits when | Stops fitting when |
|---|---|---|---|
| Single metric (day-8 ROAS) | Turns ad-spend decisions into one repeatable rule | You run paid acquisition and need a fast go/no-go signal | The metric window is too short to reflect real payback |
| Early paywall | Validates willingness to pay before building a free base | You need a product-market-fit verdict quickly | Your category only converts after long free usage |
| iOS-only launch | Keeps the discovery phase to one codebase and one funnel | Your target users concentrate on one platform | Paying users are mainly on the other platform |
| Heavy A/B testing | Surfaces which funnel elements move conversion | You have clear hypotheses and enough traffic per variant | Tests are small variations with no stated hypothesis |
Reflects the growth choices attributed to Opal's founder in a January 2024 Sub Club episode; outcomes are specific to that app and are not a guarantee for others.
| Field | What to record | Why it matters |
|---|---|---|
| Hypothesis | The single change you expect to move a named metric | Prevents tests without a reason |
| Metric | The one number that decides pass or fail | Keeps the result unambiguous |
| Window | How long the test runs before you judge it | Avoids early peeking |
| Variant | The exact change, described plainly | Makes results reproducible |
| Decision | Ship, discard, or re-test | Turns results into action |
A suggested structure, not a tool or template supplied by Opal.
Frequently asked questions
Why did Opal launch on iOS only?
According to Kenneth Schlenker, a single platform keeps the discovery phase simpler. With two platforms live you are committed to both and lose the ability to iterate quickly on one, and he cites founders who regretted launching on Android or desktop too early.
What is day-8 ROAS and why use it?
It is the return on ad spend measured over the first eight days after install. Opal used it as the one metric that told them whether paid growth was working, which keeps the decision simple: if a cohort does not return enough within that window, scaling spend is not justified.
Is an early paywall better than freemium?
They answer different questions. An early paywall tests whether people will pay now and gives a fast verdict; freemium builds a larger audience first and converts a share of it later. Opal started with a relatively expensive annual subscription and later moved toward freemium to reach new groups such as college students.
Do I need to run 121 A/B tests?
No. Schlenker's own advice is that you probably do not need that many. Fewer tests with clear hypotheses and a defined metric usually teach more per unit of effort.