I build this little vercel project to test a few things.
Project:
https://gmd-assessment.vercel.app/
Github:
https://github.com/micheleborn/gmd-assessment
The Context
This project was built as part of a Growth Marketing Designer assessment — a role that requires end-to-end ownership of landing pages, from design to deployment, with performance measured by conversion rate .
But the brief came with a twist:
- No clear audience
- No defined product positioning
- No brand guidelines
- No funnel stage
So instead of “executing,” I had to define the strategy from scratch.
The Goal
Design and build three landing page variants:
/control/variant-1/variant-2
Each with:
- A different conversion hypothesis
- A consistent structure for testing
- Measurable behavioral tracking
Step 1 — Defining the Strategy (Before Design)
Instead of jumping into UI, I started with who this is for.
Target audience (assumed)
- Mid-size DTC brands ($3M–$30M)
- Paid acquisition fatigue (Meta, CPMs rising)
- Distrust in attribution
- Pressure from finance teams
Core emotional state
“I don’t trust my growth anymore.”
That became the foundation of everything.
Step 2 — Hypothesis Design
Each variant explores a different persuasion model:
1. Control — USP Clarity
- Straightforward messaging
- “What it does” framing
- Minimal psychological layering
2. Variant 1 — Authority + Social Proof
- Borrowed credibility
- Proof-driven messaging
- Reduces skepticism
3. Variant 2 — Risk Mitigation (Most interesting)
- Focus on predictability + control
- Addresses emotional anxiety directly
- Designed for decision-makers under pressure
This is not UI variation.
This is psychological positioning testing.
Step 3 — UX & UI Decisions
UX Principles
- Single goal: conversion
- No distractions, no navigation complexity
- Repeated CTA for scanning behavior
- Clear hierarchy (problem → solution → action)
UI Decisions
- Clean, high-contrast typography
- Modular components reused across variants
- Visual consistency to isolate messaging impact
Important:
I intentionally avoided “fancy design”
Because design should not contaminate the experiment
Step 4 — Development (Next.js + Vercel)
Built with:
- Next.js + TypeScript
- Component-based architecture
- Route-based variant system
Routing strategy
/control/variant-1/variant-2
Each route:
- Injects different content
- Reuses the same components
This ensures:
- Fast iteration
- Clean experiment structure
- Scalability for future tests
Deployment
- Deployed via Vercel
- Instant previews + production parity
- CI/CD built-in
Step 5 — Analytics (PostHog)
Implemented full tracking:
Events
variant_viewedsigned_up- CTA clicks (with placement tracking)
Data captured
- Variant name
- CTA location (hero, footer, etc.)
- Referral codes
Every interaction is measurable.
Also:
- Ensured event capture before redirect
- Structured for funnel analysis
Step 6 — Experiment Design
This is the part most people skip.
What I controlled:
- Same layout across all variants
- Same CTA wording
- Same visual design
What I changed:
- Messaging strategy only
That’s how you isolate causality
What AI Changed
This project took ~3 days.
Without AI, it would realistically take ~2 weeks.
AI helped with:
- Validating messaging directions
- Exploring persuasion angles
- Speeding up iteration
But:
The strategy decisions were still human
Key Takeaways
- Conversion is psychology first, design second
- Most landing pages fail because they skip strategy
- A/B testing is useless without clear hypotheses
- Speed matters — but only if direction is correct
Final Thought
This wasn’t a design exercise.
- It was:
- Strategy
- Psychology
- Engineering
- Measurement
All in one system.
I actually love the way this turned out – super happy with the results.