7 onboarding A/B tests that move trial starts

Trial starts are the hinge metric for most subscription apps. Ads and ASO bring installs; onboarding decides how many people start a trial or hit the paywall with intent.

Not every A/B test is worth the traffic. Below are seven experiments that repeatedly move trial starts when measured on the full funnel, not isolated screens. Each includes what to change, primary metrics, and how to run it in Rheo without waiting on App Store review.

How to run these tests in Rheo

1. Duplicate your baseline flow in the builder.
2. Apply one hypothesis per variant (keep changes isolated when learning).
3. Publish each variant as its own flow version.
4. Create a channel experiment with traffic weights and a required end date (UTC).
5. Use a stable userId in the SDK for consistent bucketing.
6. At pending decision, read lift vs control on trial starts and completion; promote or extend.

The first variant in the experiment list is the control. Other arms compare pairwise at roughly 5% significance.

---

Test 1: Fewer screens before the paywall

Hypothesis: Reducing friction before the ask increases the share of users who reach the paywall and start a trial.

What to change: Remove or merge screens that repeat the same value prop. Move optional education after trial start.

Primary metrics: Paywall arrival rate, trial start rate, overall completion.

Watch out: Aggressive cuts can lower conversion if users arrive at the paywall unconvinced. Always read arrival and conditional conversion.

Control	Variant
5 screens then paywall	3 screens then paywall

---

Test 2: Social proof before the ask

Hypothesis: Trust signals immediately before the paywall increase trial starts for cold traffic.

What to change: Add or strengthen a screen with ratings, testimonial, or "join X users" copy. Keep claims truthful and store-compliant.

Primary metrics: Drop-off on the social proof screen, trial start rate, paywall dismiss rate.

Segment tip: Compare lift on paid install cohorts vs organic; social proof often helps cold users more.

---

Test 3: Paywall timing (early vs late)

Hypothesis: Showing the paywall after the user experiences a "preview" of value increases trial intent.

What to change: Variant A: paywall after welcome only. Variant B: paywall after one interactive or personalized step (quiz result, sample output).

Primary metrics: Trial starts per start, time to trial, dismiss rate.

Implementation: Two published flow versions differing only in paywall node position. RevenueCat offering stays the same.

---

Test 4: Trial-forward headline vs feature list

Hypothesis: Leading with the trial offer ("Start 7 days free") outperforms feature bullets for your audience.

What to change: On the screen before the paywall (or paywall surround copy in Rheo screens), test headline and CTA variants. Keep price terms visible.

Primary metrics: CTR on continue, trial starts, iap_purchase with period_type trial.

Compliance: Trial length and price after trial must match App Store and Play billing disclosures.

---

Test 5: Default annual vs monthly emphasis

Hypothesis: Visual emphasis on annual plan increases trial-to-paid LTV even if raw trial starts are flat.

What to change: In RevenueCat offerings, adjust default package selection or badge copy. In Rheo, test pre-paywall copy that frames annual savings vs monthly flexibility.

Primary metrics: Trial starts, purchased product id distribution, 30-day retention (RevenueCat).

Note: Product selection UI is RevenueCat's; Rheo experiments on the path and framing into that UI.

---

Test 6: Permission prompts before vs after paywall

Hypothesis: Deferring ATT or push permission until after trial start improves trial volume without destroying opt-in rates.

What to change: Control: ATT pre-prompt before paywall. Variant: paywall first, then ATT on premium welcome.

Primary metrics: Trial starts, ATT opt-in rate, notification opt-in if applicable.

Platform: iOS ATT timing must follow Apple guidelines. Rheo surfaces builder hints for attribution-sensitive flows.

---

Test 7: Second chance after paywall dismiss

Hypothesis: Users who dismiss the first paywall will start a trial when offered a softer follow-up.

What to change: Wire RevenueCat dismissed to a Fallback path: one screen with a reduced ask, reminder of trial, or feature teaser, then optional second paywall presentation.

Primary metrics: Trial starts per user (including second touch), dismiss rate on second presentation.

Caution: Do not nag. One structured second chance beats repeated modal loops.

---

Summary table

#	Test	Main lever	Key metric
1	Shorter path	Screen count	Trial starts / start
2	Social proof	Trust screen	Trial starts
3	Paywall timing	Node order	Conditional conversion
4	Trial-forward copy	Headline / CTA	Trial starts
5	Plan emphasis	Framing + RC offering	Trial + product mix
6	Permission order	ATT / push position	Trials + opt-in
7	Second chance	Dismiss branch	Incremental trials

Experiment hygiene

• One big change per variant when possible. Otherwise you will not know what worked.
• Plan sample size before launch. Extend at pending decision rather than peeking daily and stopping early.
• Filter analytics by experiment variant and published version so pre/post pins do not blend.
• Watch sample ratio mismatch. Uneven splits invalidate comparisons.

Test 7 (continued): Measuring incremental trials

For second-chance paths, define incremental trials as purchases among users who already dismissed once. Compare variant with Fallback enabled vs control with hard stop on dismiss. Incremental lift can justify softer UX without annoying users who would have converted on first presentation.

---

Cross-vertical notes

Vertical	Test that often wins first
Productivity	Test 3 (preview before paywall)
Health / fitness	Test 2 (social proof)
Education	Test 1 (shorter path)
Finance	Test 4 (clear trial terms)

Verticals differ; let your funnel cliffs override defaults.

Prioritizing which test to run first

Use your funnel analytics to choose, not a generic playbook:

If your data shows…	Start with test
Cliff on screen 2-3	2 (social proof) or 4 (copy)
High paywall arrival, low purchase	4, 5, or 7
Low paywall arrival	1 or 3
iOS ATT opt-in under 25%	6
Many dismiss, few second touches	7

Export a screenshot of the funnel chart into your experiment doc so stakeholders agree on the bottleneck.

Sample size reminder

Trial start rate differences of 2-3 absolute points need large samples. Use Rheo's experiment sample planning while still in draft. If you hit pending decision underpowered, extend the end date instead of calling a winner from noise.

Approximate baseline trial rate	Detectable lift (relative)	Order of magnitude starts per arm
5%	20% relative (~1 pt absolute)	Tens of thousands
10%	15% relative	Thousands to tens of thousands
20%	10% relative	Thousands

Low-traffic apps should run fewer concurrent tests and longer windows.

Documenting results

For each concluded test, record:

• Hypothesis and primary metric
• Control and variant screenshots
• Start/end dates and environment
• Decision: promote, extend, or inconclusive
• Follow-up test queued

This prevents re-running the same failed idea six months later because someone left the team.

What not to test (yet)

• Micro color tweaks before you have traffic for power.
• Ten variants at once (multi-arm noise inflates false positives).
• Pricing changes hidden from store listings.
• Paywall product mix and onboarding length in the same variant.

Fix the structural tests above first. They compound.

Ship the next test this week

Pick the test closest to your current funnel bottleneck. If drop-off is heavy on screen 2, try Test 2 or 4. If many users reach the paywall but dismiss, try Test 7. Publish variants OTA, start the experiment, and read results where onboarding and billing meet.

Building a quarterly test roadmap

Month	Focus
Month 1	Path length (Test 1) + diagnosis
Month 2	Paywall timing (Test 3)
Month 3	Permission order (Test 6) or second chance (Test 7)

Sequential tests on the current winner keep interaction effects manageable while maintaining velocity.

Pairing with step-level analytics

Before each test, screenshot the funnel and drop-off panels for the control pin. After promote, compare the same panels on the winner. Tests that move trial starts without moving the pre-paywall cliff often indicate paywall-only luck; tests that smooth the cliff usually sustain gains longer.

Growth is a portfolio of small bets. These seven tests are the bets that most often pay back on subscription apps when you measure them on the full path from install to trial.

Start for free