Executive Summary

Problem Legacy funnel builder causing enterprise churn—slow, error-prone, and impossible to visualize
Solution Inline editing flow with auto-generated pages and error-proof integrations
Key Result +37% Commercial Value Lift
+37% Commercial Value Lift

Upsell Funnel Builder Redesign

Complete redesign of a revenue-critical feature that stopped enterprise churn and drove 37% lift in commercial value.

Role
Lead Designer PM, 2 Engineers, Data Analyst
Timeline
3 Months No dedicated sprint—parallel to roadmap
Impact
+37% Revenue 0 enterprise churn, +40% conversion

The Problem

The Upsell Funnel Builder drives significant revenue by enabling creators to sell additional products post-checkout. But our enterprise customers were churning because of it.

Through support ticket analysis and user interviews, I identified four core issues:

1

Painfully Slow

Creating a funnel required 3-4 modal windows per step, plus manual button configuration on every page. Power users spent 15+ minutes on simple funnels.

2

Constant Context Loss

Tree view existed but was constantly interrupted by modal windows. Every action broke focus, forcing users to mentally rebuild context after each step.

3

Error-Prone Integrations

Custom website integration had scripts and buttons in wrong order. Users consistently placed code incorrectly, breaking funnels and flooding support.

4

No Product-Level Editing

Users had to leave product settings and navigate to a separate feature area to edit funnels. Constant context switching slowed workflows and caused frustration.

User Journey Map
Journey mapping revealed friction at every stage of funnel creation

The Solution

I redesigned around four principles: make it fast, make it visible, make it foolproof, and keep users in context.

1. Speed — From 15 minutes to 2 minutes

Eliminated modal windows entirely. Replaced with inline side panel editing and auto-generated funnel pages with BUY/DON'T BUY buttons pre-configured.

Before — Modal-heavy, manual setup
After — Inline editing, auto-generated pages

2. Clarity — Tree view that actually works

The tree view existed, but constant modal interruptions broke focus. Now users stay in the tree—edit inline, see changes instantly, and delete entire branches with one click.

Before — Multiple modals to add a node
After — Click, configure in side panel, done

3. Error Prevention — Designing out the #1 support issue

Reordered the custom website integration UI to show scripts and buttons in correct implementation order, with explicit instructions. Also added visibility into which products connect to each funnel.

Before — Wrong order caused broken funnels
After — Correct order with clear guidance

4. Context — Edit where you work

Previously, editing a funnel required leaving product settings entirely. Now users can edit funnels directly from the product area—no more jumping between sections.

Before — Preview only, had to leave product area
After — Full editing from product settings

Validation

Before development, I tested prototypes with 10 power users. Key findings:

Inline editing was the unlock

Users loved staying in the tree. "I can finally see what I'm building without losing my place."

Auto-generation exceeded expectations

Pre-configured pages with BUY/DON'T BUY buttons removed 80% of setup time. Users didn't miss the customization.

Product-level editing was table stakes

Users expected to edit funnels where they assigned them. The old workflow felt broken to every tester.

Impact

+37% Commercial Value Lift

Average funnel revenue per user increased significantly

0 Enterprise Churn

Zero accounts churned citing funnel issues post-launch

+40% Order Conversion Efficiency

More orders completed per funnel created

Results Chart
Performance metrics before and after the redesign

Reflections

1

The tree wasn't broken—the modals were

We initially explored alternative visualizations. Turned out the tree worked fine—we just needed to stop interrupting it with modals.

2

Ship incrementally next time

We shipped everything at once. Feature flags would have let us validate inline editing independently before adding auto-generation.