About Heatseeker

Heatseeker is an AI-powered experiment platform that helps teams validate demand using stealth ads and synthetic personas so they can test ideas before building

Introduction

Heatseeker’s trial users had an activation cliff: 83% of new trial users never launched their first experiment. I redesigned sign-in → onboarding → dashboard → “Start experiment” into a guided, AI-assisted workflow designed to get first experiment launch under 90 seconds

Overview

Heatseeker’s “secret sauce” combines stealth ads (live demand signals) + synthetic personas (predict scale). Powerful, but intimidating for first-time users who just want a fast A/B style test

The problem

A mission-critical drop-off: 83% of trial users never launched their first experiment , despite the product promise of “real market evidence in minutes”

Goal

Help a first-time user sign in, move through onboarding, and launch the first experiment in < 90 seconds.
Scope: Sign-in → Onboarding → Dashboard → Start Experiment flow

Persona (what guided the tradeoffs)

“Launch-Hungry Sara”

• Comfortable with SaaS dashboards

• Proficient with common UX patterns

• Ultra-impatient: bounces if value isn’t obvious in < 2 minutes

What the legacy experience forced users to do

1. Sign in (SSO or email/password)

2. Choose what they’re testing + audience + goal

3. Generate & review experiment

4. Set budget

5. Simulate / launch

Design exploration (3 directions → 1 shipped)

Option 1 - Safe: reduce noise

User problem

“I’m confused. Where do I look? Too much info and no clear start”

Intent

Simplify dashboard hierarchy + clarify primary CTA

Essential (Safe Option)

Option 2 - Stretch: goal-centric workspace

User Problem

“I don’t know my goal / am I testing the right thing?”

Intent

Reframe the dashboard around goals + metrics that matter

Goal-Centric (Stretch Option)

Option 3 - Wild: AI growth copilot

User problem

“I don’t know what to test next, I just want results fast”

Intent

Move from “tool UI” to guided outcomes, where AI reduces setup decisions

AI-Powered Assistant (Wild Option)

Final direction (what I designed)

1) Two dashboards: New users vs power users

• New users: guided “next best action” to reach first launch fast

• Power users: dense control + history without slowing them down

Dashboard For New Users

Dashboard For New Users

Dashboard For Power Users

Dashboard For Power Users

2) AI as a collaborator, not a black box

The UI supports both paths:

1. AI-assisted: faster start, fewer decisions

2. Manual: power users can skip straight to control

Create Experiment Flow (Using AI & Manually)

Created with AI Results

Created with AI Results

Simulating Experiment

Simulating Experiment

What I’d do next

• Instrument the funnel end-to-end (step drop-offs, time to first launch, edit rate after AI generation)

• Add “What to test next” recommendations post-launch to create a repeatable experimentation loop

Recruiter Testimonial