Step-level funnel analytics: finding the screen that kills trial starts

Your paywall dashboard shows views, trials, and conversions. It does not show why only 12% of installs ever reach the paywall. When trial starts stall, teams rewrite paywall copy by default. Often the leak is three screens earlier: a permissions wall, a confusing value prop, or a branch that sends half your cohort into a dead end.

Step-level funnel analytics maps every screen in the journey, measures drop-off between steps, and ties those steps to revenue outcomes. This is how you find the screen that kills trial starts before you waste another sprint on the wrong hypothesis.

The limits of paywall-only metrics

Paywall tools and billing SDKs excel at monetization events: offerings shown, trials started, renewals, refunds. That is necessary but incomplete for funnel optimization.

Paywall-only view	What you miss
Trial conversion rate	Users who never reached the paywall
ARPU by cohort	Which onboarding path produced paying users
Offering performance	Screen order and messaging that created intent
Experiment lift	Whether variant B won on step 5 or only at checkout

Consider a simple install → trial funnel:

Install (100%)
  → Onboarding start (72%)
    → Paywall view (31%)
      → Trial start (9%)

A 9% install-to-trial rate might look like a paywall problem. But 41% of users quit onboarding before they ever see pricing. Fixing the paywall alone cannot recover them.

Step-level analytics decomposes the 72% → 31% collapse screen by screen so you fix the binding constraint.

Define steps consistently

Before you analyze, standardize how steps are named and fired.

Use stable step IDs. Screen titles change during experiments. onboarding_step_04_permissions should persist across copy tests.

Fire events at render, not only on button tap. You need to know who saw a screen, not only who acted.

Record variant assignment at first exposure. Every downstream step should be attributable to control or treatment.

Align with revenue events. Trial start, purchase, and refund events from RevenueCat or native billing should share user and session identifiers with funnel events.

In Rheo, each screen in a published flow automatically emits step events to analytics. Experiments attach variant labels so funnels split by treatment without extra instrumentation per screen.

Example funnel walkthrough

Below is a fictional but realistic onboarding funnel for a subscription productivity app. Numbers are illustrative.

| Step | Screen | Users entering | Step conversion | Cumulative from install | |------|--------|----------------|-------------------------| | 1 | Welcome | 10,000 | 94% | 94% | | 2 | Goal selection | 9,400 | 88% | 83% | | 3 | Feature highlights | 8,272 | 76% | 63% | | 4 | Account creation | 6,287 | 52% | 33% | | 5 | Paywall | 3,269 | 41% | 13% | | 6 | Trial confirmed | 1,340 | n/a | 13% |

Reading the table:

• Steps 1–3 lose users gradually (expected).
• Step 4 loses 48% in one hop. This is the anomaly.
• Step 5 paywall conversion (41% trial start among viewers) is actually reasonable.

The business sees "low trial starts" and requests new paywall creative. The data says: unblock account creation first.

The "screen 4" problem pattern

In dozens of onboarding reviews, the highest single-step drop often clusters around:

1. Account creation walls before value is shown.
2. Permission stacks (notifications + ATT + location in one sequence).
3. Surprise paywalls with no price context earlier in the flow.
4. Broken branches where one choice leads to an empty state or loop.

We call this the screen 4 problem because it is rarely the first or last screen. Early screens benefit from novelty. The paywall gets explicit optimization attention. Middle screens accumulate legacy requirements from legal, marketing, and engineering until they become a choke point.

Diagnostic questions for any high-drop step:

• Did users who dropped see a loading state longer than 3 seconds?
• Is the step skippable for motivated users?
• Does one acquisition channel drop more than others?
• Did a recent copy change correlate with the cliff?

Segment drops before you fix

Aggregate funnels hide actionable segments.

By acquisition source: Instagram ads may tolerate longer quiz flows. Brand search users may want faster time to price.

By platform: iOS and Android often diverge on permission steps and payment friction.

By geo: Localization gaps show up as cliffs on specific screens, not globally.

By variant: In an A/B test, compare step conversion, not only final trial rate. A variant can win at the paywall while losing users on step 3.

Segment	Step 4 drop	Insight
iOS, US, paid social	51%	Account wall after short flow
iOS, US, organic	29%	Motivated users tolerate signup
Android, IN	62%	Phone verification fails often

A global "simplify step 4" test might help paid social on iOS while hurting organic quality. Segment first, then design targeted branches or experiments.

From insight to experiment

Step-level data should end in a test plan, not a slide deck.

Workflow:

1. Identify the step with the largest absolute drop (users lost, not only rate).
2. Formulate a hypothesis tied to that step ("Reducing fields on account creation will increase paywall reach").
3. Choose a primary metric (users reaching paywall) and guardrails (trial-to-paid rate).
4. Ship a variant remotely when possible so you are not blocked on App Store review.
5. Read step metrics again before calling a winner.

For the fictional app above, a sensible experiment backlog:

Test	Change	Primary metric
A	Defer account creation until after paywall trial start	% reaching paywall
B	Social sign-in only on step 4	Step 4 conversion
C	Show price teaser on step 3	Step 5 trial conversion

Run A first. If paywall reach jumps but trial quality drops, segment by channel or add qualification earlier for low-intent traffic only.

Tie steps to revenue, not just completion

Completion rate is a means. Revenue is the end.

For each funnel step, track downstream value:

• Trial start rate among users who passed step N.
• Trial-to-paid by onboarding path.
• ARPU at 30 days by variant.

A variant that increases paywall views by 15% but decreases trial-to-paid by 20% is a net loss. Step funnels let you see where quality degrades, not only where quantity moves.

Rheo connects step analytics to experiment results so you read lift on trials, purchases, and revenue in one place alongside per-screen drop-off.

Tooling checklist

Whether you use Rheo or assemble your own stack, you need:

• Per-step enter and complete events
• Variant / experiment assignment persisted per user
• Billing event integration (RevenueCat, etc.)
• Cohort date filtering aligned to experiment windows
• Export or API for weekly review rituals

Avoid dashboards that only show daily active users and purchase totals. They cannot localize onboarding leaks.

Weekly review template

Block 30 minutes every Monday:

1. Export step funnel for the last 7 days.
2. Compare to prior week and to experiment cohort if live.
3. Note the single largest absolute drop.
4. Assign one owner to the next hypothesis.

Consistency beats sophisticated tooling nobody opens.

Connecting experiments to step data

When you launch an onboarding A/B test, compare step conversion deltas, not only final trial rate. A winning variant might improve step 4→5 by 8% while step 5→paywall is flat. That tells you the change helped mid-funnel motivation, not paywall framing.

Hold a short post-experiment review: screenshot the funnel table, note variant ID, archive in Notion or your experiment log. Future you will forget why August's test was reverted.

RevenueCat and billing alignment

When trials start through RevenueCat, map billing events back to funnel steps. A user who starts trial on step 6 should carry step history in your analytics warehouse for LTV analysis. Without that join, you optimize screens blind to paid conversion quality.

FAQ

How is this different from product analytics funnels in Mixpanel or Amplitude?
Conceptually similar. The difference is tying funnels directly to remote flow versions and experiments without redeploying the app for each change. Flow-native analytics keeps step definitions aligned with what is live.

How many steps should we track?
Every screen in the guided onboarding and paywall path. Branches should appear as separate paths, not folded into one average.

What if drop-off is even across all steps?
Look outside onboarding: app crashes, load time, ad mismatch, or broken deep links. The problem may be pre-flow.

Find your leak, then fix it

Trial starts improve when the right users reach the paywall with intent. Step-level funnel analytics shows where intent is lost and whether your experiments recover it.

Start for free. Map your onboarding on the Rheo canvas, publish over the air, and read the step that actually kills trial starts.