Puzzle Onboarding funnel

The Challenge

Puzzle's AI-automated bookkeeping was technically differentiated, but invisible until users were already inside the product. The existing onboarding asked too much too late, causing drop-off before activation. The paywall felt punitive rather than earned, and high-intent accounts weren't being captured efficiently.

My Role

Lead Product Designer - Led the discovery work of understanding the existing funnel flow - dropoff points, understanding which user segments; translated findings into prototyped wireframe flows for testing and research; UI, interaction and system design.

Tools used - PostHog, Figjam, Figma, Cursor, Claude Code, Vercel, Superbase

The Process

Reframing the funnel as a qualification engine. Every question was designed to serve double duty — configure the product intelligently and signal the right pricing tier. That reframe turned a compliance task into a product strategy.

Progressive disclosure over 7 steps keeps cognitive load low. Sign-up comes first with three auth paths (email, Google, Rippling SSO), followed by personal context before company setup — letting Puzzle segment by role before ever touching account configuration.

Steps 2–6 use radio pill selectors throughout, with microcopy that earns trust: "Don't worry, you can always change this later" and "We use this to..." justifications on questions that would otherwise cause hesitation. Step 6 is the only field marked optional, reducing abandonment at the finish line.

The paywall rewards rather than interrupts. It appears as a modal overlaid on a blurred dashboard — users see what they're unlocking. Four tiered plans ($25–$300/mo), a "Most popular" callout with a distinct CTA, and trust badges (14-day trial, cancel anytime) reduce commitment friction. A monthly/annual toggle creates a natural upsell moment at peak intent.

The Outcome

The funnel reduced perceived friction, improved user qualification, and reframed the paywall as a natural conclusion to setup. Critically, the data collected (entity type, industry, revenue model, expenses) maps directly to auto-configuring the chart of accounts — turning the conversion funnel into a personalization engine that compresses time-to-value. The fully component-driven architecture also enables fast A/B testing of individual steps without accruing design debt.

Interactive prototype

Built with Cursor and Claude Code, using Next.JS and deployed with Vercel.