Hero Innovation Group

What the engagement looked like.

Hero had a product in production, a paying customer base, and an engineering process that had stopped scaling. The team wanted the same product — faster, cleaner, and visually owned in-house — without paying the cost of a stop-the-world rewrite. I joined to run the parallel track.

The problem on arrival.

The existing codebase had accumulated patterns that worked against the people writing them: conventions were implicit, the design system lived in a contractor’s Figma file, and new features took longer each quarter.

Visual direction was outsourced to an agency whose output didn’t survive contact with the real product — every handoff needed translation. Whoever was going to rebuild had to also absorb the design role.

How it came together.

4 phases, each one trying to prove a single thing before the next could start.

1. Month 1
   Parallel scaffold

   New codebase stood up alongside the production one. Shared schema, separate conventions. No pressure to feature-match on week one — only to prove the patterns.

2. Month 2
   Design brought in-house

   Pulled the visual system out of the contractor Figma into a lightweight token set the team could edit. Components redrawn where the agency version couldn’t survive real data.

3. Month 3
   Vertical catch-up

   Every feature in the production app shipped in the new one, one vertical slice at a time. Each slice was a chance to pick a cleaner pattern and keep it.

4. Month 4
   Cutover

   Users migrated without a downtime window. The old app stayed around a few weeks as read-only fallback, then retired.

What shipped.

The rebuilt product took over from the old one inside a quarter, with the visual direction now owned in-house and a codebase the team could actually extend. Two product lines ran in parallel for most of the engagement — one shipping, one being replaced — without either stalling.