Fixing Funnel Drop-Off with UX

Session Recording #47 Made Me Want to Redesign Everything

We'd been looking at funnel metrics for weeks. Conversion step 3 had a 58% drop-off. The analytics said "users leave at step 3." That was true and completely useless — it told us where but not why.

I sat down with 200 session recordings across the funnel. By recording 47, I'd stopped taking notes and was just watching with a knot in my stomach. A user spent 4 minutes on step 3. They filled in a field. Then deleted it. Filled it in again. Scrolled up, scrolled down. Looked for something that wasn't there. Then left.

We weren't losing users because they didn't want to complete the flow. We were losing them because the flow was making them feel stupid. They couldn't find what they needed to answer the question we were asking.

That was the moment the optimization playbook changed — from "reduce friction" to "remove confusion." They're different problems.

The Funnel Data

Where Everyone Was Leaving

Funnel Step	Completion Rate	Drop-off Rate	Avg Time on Step
Step 1: Basic Info	89%	11%	45 sec
Step 2: Contact Details	81%	19%	1 min 20 sec
Step 3: Policy Details	42%	58%	4 min 10 sec
Step 4: Payment	71%	29%	2 min 45 sec
Step 5: Confirmation	94%	6%	30 sec

Step 3 had a 4-minute average time — 5× longer than Step 1 — and a 58% drop rate. Users weren't abandoning quickly. They were struggling for 4 minutes and then giving up. That time difference was the tell.

Standard analytics would have said "optimize step 3." Session recordings showed us what that optimization needed to be: users couldn't answer the question "Vehicle Registration Number" without knowing what it looked like on their documents. We'd provided no hint, example, or format guidance.

What the Recordings Showed

200 Recordings. 6 Failure Patterns.

I tagged every recording with the primary reason for exit or struggle. Six patterns appeared repeatedly. Together they explained over 85% of step 3 abandonment.

Failure Pattern	% of Abandoners	Root Cause
Can't find required document	31%	Field asks for document number; no guidance on which document or where to find it
Unsure what format is correct	22%	Validation error appears but doesn't explain correct format
Confused by insurance jargon	18%	'IDV', 'NCB', 'Zero Depreciation' — no tooltips, no plain-language explanation
Mobile keyboard covering form fields	8%	Last 2 fields of step 3 hidden behind keyboard; no scroll compensation
Distrust of data being collected	4%	Asking for policy number before trust established
Other / unclear	17%	Various individual issues

The #1 failure was a document lookup problem, not a form design problem. Users couldn't complete the form not because of friction but because they physically didn't have the information in front of them. The solution wasn't a better form — it was not asking for that information at that stage.

The Design Shift

From "Reduce Friction" to "Remove Confusion"

Traditional funnel optimization focuses on reducing friction: fewer fields, shorter forms, fewer steps. That wasn't the right frame here. The issue wasn't effort — users were clearly willing to put in effort (4 minutes average). The issue was comprehension and confidence.

When users spend 4 minutes on a step and then leave, they didn't find the step too hard. They found it impossible. The difference between "difficult" and "impossible" is whether the user has what they need to complete the task.

The redesign principle: every field in the form needed to be answerable by someone sitting on their couch with their phone — without needing to go find a document they may not know exists.

The Redesign

What We Changed and Why

Progressive Disclosure

Vehicle registration moved from step 3 to an optional enhancement step post-purchase. Quote can be generated with just make/model/year. Pre-fill from registration happens if user provides it — but it's no longer a gate.

Field-Level Context

Every field with potential confusion now has: a plain-language label, a tooltip with 'where to find this,' and a format example. 'NCB' became 'No Claim Bonus (years without a claim).' Validation errors show correct format, not just 'invalid.'

Mobile Scroll Compensation

Form auto-scrolls to keep the active field above the keyboard on iOS and Android. Previously, fields 4 and 5 of step 3 were obscured by the keyboard — users typed blind.

Jargon Elimination

Every insurance term used in the form audited and replaced with plain language. Where jargon was unavoidable, inline explanation added. 'IDV' → 'Insured Declared Value (what your insurer pays if your vehicle is totalled).'

Implementation

What Was Built

Form Architecture Redesign

→Step 3 split into 3a (must-have) and 3b (nice-to-have / post-purchase)
→Vehicle registration moved to optional enhancement step
→Field order reorganized: easy fields first, document-required fields last (or deferred)

Contextual Help System

→Tooltip component built: shows on field focus, not hover (mobile-first)
→Format examples: 'MH02 AB 1234' for registration, '2 years' for NCB
→Validation errors rewritten: 'Should look like MH02 AB 1234' instead of 'Invalid format'

Mobile Keyboard Handling

→Active field scroll-into-view implemented across iOS and Android
→Form container height adjusted dynamically when keyboard visible
→Submit CTA always visible above keyboard on mobile (floating, not inline)

Results

Testing and Results

Week 1–2 · Session Recording Analysis

Diagnosis

→ Clear redesign hypothesis: progressive disclosure + field context + mobile fix.

✓200 recordings tagged with primary failure reason
✓6 failure patterns identified and quantified
✓Redesign brief written based on failure pattern data

→ Clear redesign hypothesis: progressive disclosure + field context + mobile fix.

Week 3–5 · Build

Implementation

→ Staged on test environment. Internal review completed.

✓Form architecture redesign: registration deferral
✓Tooltip system built
✓Mobile keyboard handling implemented

→ Staged on test environment. Internal review completed.

Week 6–8 · A/B Test

50/50 Split

→ Full funnel completion: 26% → 47%. Revenue per 1,000 visitors: +81%.

✓Old flow vs new flow: equal traffic split for 3 weeks
✓Statistical significance reached at Week 6
✓Step 3 completion: 42% → 74%

→ Full funnel completion: 26% → 47%. Revenue per 1,000 visitors: +81%.

+76%

Step 3 Completion

42% → 74%

+81%

Revenue per 1K Visitors

full funnel impact

−62%

Avg Step 3 Time

4 min → 1.5 min

+34%

Mobile Completion

biggest gain channel

Insights

What I Now Believe About Funnel Optimization

Time on step is the most underused diagnostic metric

High time + high drop = confusion, not laziness. Low time + high drop = too hard (users give up immediately). High time + low drop = high engagement. Reading time alongside drop rate tells you what kind of problem you're solving.

Session recordings are worth 10× funnel analytics for diagnosis

Funnel analytics tells you where. Session recordings tell you why. The 4-minute struggle visible in Recording #47 was invisible in the drop-off percentage. You need both — analytics to prioritize, recordings to diagnose.

'Remove friction' and 'remove confusion' are different problems

Friction reduction = fewer steps, fewer fields. Confusion removal = better context, clearer labels, appropriate field timing. Applying friction-reduction solutions to confusion problems will improve metrics without solving the actual user experience issue.

Mobile keyboard handling is chronically underinvested

8% of step 3 failures were from fields hidden behind the mobile keyboard — entirely invisible in any desktop QA process. Separate mobile-first testing and keyboard-visible flow review should be standard for any form.

Failures

What Didn't Work First

✕ First redesign just added progress indicators

Why it broke

Initial hypothesis was that users didn't know how long the form was. Added a step progress bar. Drop rate moved from 58% to 54%. Improvement, but not where the real problem was.

What fixed it

Session recordings showed users who could see the progress bar still left at the same point in step 3 — when they couldn't find their vehicle registration. Progress visibility wasn't the issue; answerable questions were.

✕ Tooltip content was too technical

Why it broke

First version of tooltips used insurance industry standard language. 'Registration certificate (RC)' — but users called it 'RC book' or 'papers' and didn't recognize 'registration certificate.'

What fixed it

Rewrote tooltip content using the language users used in session recordings and support chats. 'Your RC book — the green booklet from the RTO.' Recognition shot up.

Future

What's Next

Continuous Recording Analysis

Set up automated session recording tagging using heuristics (rage clicks, u-turns, long pauses) to surface struggling sessions automatically — rather than manual review.

Predictive Field Completion

Pre-populate fields from known user data (previous policies, partner integrations, device-based autofill). Reduce required manual input — the best field is the one the user never had to type.

Per-Field Analytics

Track time, error rate, and abandonment at the individual field level (not just step level). Identify which specific fields are causing struggle before they affect step completion rates.

The funnel didn't need fewer steps. It needed to stop asking users questions they couldn't answer from their couch. Fix that, and the drop-off fixed itself.

Designing a Growth Loop That Brought 30% Repeat Users in 14 Days

Growth

How We Built a Zero-Cost Growth Engine Using Organic + Product Hooks

Growth

1200% Insurance Revenue Growth in 12 Months: The Full Funnel Story

13 min read

Growth

From 3.8M to 45M+ Monthly Active Users: The CarInfo Growth Architecture

14 min read

Growth

How We Cut User Acquisition Cost by 30% Without Cutting Volume

11 min read

All Case Studies

Browse all 30 breakdowns

Work With Me

Let's build something together

Connect on LinkedIn

Follow for more insights

Where to go next

Product Monetisation Consultant
The consulting work this sits inside, with the rest of the evidence behind it.
Average revenue per user benchmark, subscription apps
Published figures, traced to their sources — and where the sources disagree.
Trial-to-paid conversion benchmark, subscription apps
Published figures, traced to their sources — and where the sources disagree.
Product Growth Score
Thirteen questions that locate which part of the funnel is actually constraining growth.

Optimizing Funnel Drop-Off Using Micro UX Interventions

Results at a glance

Session Recording #47 Made Me Want to Redesign Everything

Where Everyone Was Leaving

200 Recordings. 6 Failure Patterns.

From "Reduce Friction" to "Remove Confusion"

What We Changed and Why

What Was Built

Testing and Results

What I Now Believe About Funnel Optimization

Time on step is the most underused diagnostic metric

Session recordings are worth 10× funnel analytics for diagnosis

'Remove friction' and 'remove confusion' are different problems

Mobile keyboard handling is chronically underinvested

What Didn't Work First

What's Next

You might also like

Where to go next

Session Recording #47 Made Me Want to Redesign Everything

Where Everyone Was Leaving

200 Recordings. 6 Failure Patterns.

From "Reduce Friction" to "Remove Confusion"

What We Changed and Why

What Was Built

Testing and Results

What I Now Believe About Funnel Optimization

Time on step is the most underused diagnostic metric

Session recordings are worth 10× funnel analytics for diagnosis

'Remove friction' and 'remove confusion' are different problems

Mobile keyboard handling is chronically underinvested

What Didn't Work First

What's Next