
Run the same onboarding experiment on web and mobile
Design cross-platform onboarding experiments with shared metrics, variant assignment, and platform-aware rendering in Rheo Convert for web and mobile apps.
Web signups and mobile installs are not two products: they are two doors into the same subscription business. Yet most teams run separate onboarding codebases, separate analytics projects, and separate experiment consoles. Leadership asks “which variant won?” and engineering answers with two incompatible spreadsheets.
Rheo Convert lets you run one experiment across web and mobile: shared hypothesis, shared variant keys, shared funnel metrics: while each platform renders native UI. This post covers setup, metric alignment, platform rendering caveats, and how Analytics rolls up results without double-counting users.
Why cross-platform experiments matter
| Scenario | Risk of siloed tests |
|---|---|
| Web ads drive mobile install | Web winner hurts app completion |
| Single account web + app | User sees inconsistent journeys |
| Shared pricing story | Paywall copy diverges by platform |
| Portfolio reporting | Cannot rank one backlog |
One experiment definition forces explicit tradeoffs: “We accept shorter web path; mobile keeps permission education.”
Architecture: one flow graph, multiple channels
In Rheo:
- Flow holds screens, logic, and experiment definition.
- Channels bind the flow to surfaces: iOS app, Android app, web URL or embedded web.
- Renderer paints platform-appropriate components (SwiftUI, React Native, web).
- Analytics records
flow_started, step events, outcomes withflowId,versionId,variantId, and platform dimension.
Users on web and mobile contribute to the same experiment if they enter through channels attached to that experiment’s flow version.
Step 1: Align the hypothesis and primary metric
Before building UI, write:
| Field | Example |
|---|---|
| Hypothesis | Removing one carousel screen increases trial start rate without hurting paid conversion |
| Primary metric | trial_started within 24h of flow_started |
| Guardrail | purchase_completed / revenue per started user |
| Minimum runtime | 7 days or powered sample (see experiment planning docs) |
Shared metric does not mean identical UI: means the event names and success criteria match.
Step 2: Design for platform parity (where it matters)
| Element | Parity approach |
|---|---|
| Copy and value prop | Same strings; watch length on small screens |
| Step count | Same logical steps; optional platform-only permission screens |
| Paywall offer | Same products via RevenueCat / Stripe |
| Media | Responsive assets; web may use video webm, mobile mp4 |
| Navigation | Swipe vs click: renderer handles affordances |
Document intentional divergences in the experiment readme: “Android includes notification pre-prompt; web skips.”
Analyze with platform segment so Android-only steps do not look like drop-off on iOS.
Step 3: Variant assignment consistency
Users should stick to one variant per experiment:
- Assignment keyed on Rheo customer ID after
identify. - Anonymous web visitors get anonymous ID; merge on login when possible.
- Avoid clearing storage in ways that re-roll assignment mid-funnel.
Cross-device login: if user starts on web and continues in app, merged profile should retain variant: test this path in QA.
Step 4: Shared analytics, platform cuts
In Analytics funnel view:
| View | Use |
|---|---|
| All platforms | Headline experiment result |
| iOS / Android / Web | Detect renderer-specific cliffs |
| Locale | Copy overflow issues |
| Pitfall | Fix |
|---|---|
Web flow_started without mobile | Weighted reporting by traffic share |
| Mobile crash on one OS | Filter by app_version |
| Different entry rates | Normalize by channel traffic, not raw counts |
Compare to step-level funnel analytics for trial starts for reading cliffs.
Step 5: Ship without synchronized app store releases
Web channel updates on publish pin immediately. Mobile requires SDK that loads remote flow: no store release for copy/layout changes inside the Rheo graph.
Store release still needed for new native layer types or SDK upgrades: not for experiment copy swaps.
Platform rendering notes
Web
- Respect analytics and marketing consent banners before SDK init where required.
- Deep links from Engage email should land on web channel with UTM preserved for attribution.
React Native / Expo
- Test safe areas, keyboard overlap on forms, and WebView paywall nodes if used.
- OTA limits: flow JSON is remote; binary must include renderer for layer kinds in graph.
SwiftUI
- Simulator and device QA for dynamic type and dark mode.
- Permission flows differ from web: branch explicitly.
Shared paywall nodes
When using RevenueCat or Superwall nodes, confirm same offering IDs across platforms. Billing UI may look different; purchase events must map to same analytics outcomes.
Engage follow-up on the same experiment
Users who abandon web onboarding can enter Engage automations with deep links back to the same channel: variant preserved if profile merge works.
See from funnel drop-off to Engage campaign.
Experiment checklist
- Primary metric event fires on all platforms in staging
- Three QA paths: web only, mobile only, web → mobile login
- Platform segment dashboards saved
- Consent gates verified per platform
- Rollback pin identified before starting test
- Stakeholders agree platform-specific cliffs are acceptable
When not to force one experiment
| Case | Alternative |
|---|---|
| Web is marketing site only; app is product | Separate flows; link attribution only |
| Radically different SKUs web vs mobile | Separate paywall experiments |
| Immature web renderer | Mobile-first experiment; web copy-only |
Statistical and traffic considerations
Cross-platform tests need enough entrants per variant per platform to read subgroups. If web contributes 5% of traffic, web-only cliffs may stay noisy for months.
| Strategy | When |
|---|---|
| Pooled analysis for ship decision | Primary metric lifts overall with guardrails met |
| Platform-specific follow-up | Subgroup shows harm after pooled win |
| Sequential test | Ship mobile winner, then adapt web |
Document minimum detectable effect before launch. Rheo experiment tooling pairs with sample planning in product docs: do not run two-week tests on 200 total users expecting definitive revenue conclusions.
QA matrix (copy into your test plan)
| Path | Device | Expected variant stickiness |
|---|---|---|
| Cold web signup | Desktop Chrome | Yes |
| Mobile web Safari | iPhone | Yes |
| App install after web account | iOS app | Merged profile keeps variant |
| Logout / login | Both | No re-roll |
| Deep link from Engage email | Web channel | Lands in correct step |
Failed merge paths are the top source of “experiment says tie, product says chaos.”
Bottom line
Cross-platform growth needs one experiment identity and honest platform slices. Rheo Convert plus Analytics gives shared variants and funnels; your job is to document where native reality diverges: and read the cuts before calling a winner.