When "Almost Done" Isn't

This project had a reputation

Checkout 2.0 started as a legitimate business problem. The checkout flow was so simple that the care center was getting buried in low-intent leads that never converted. The redesign was supposed to fix that. Get customers to fail early, qualify for financing first, and know the full scope of what they’re walking into.

What followed was two years of design, redesign, and handoffs with nothing to show for it. Development had already started. The project had passed through two teams. One person quit over it. Another was fired. The business kept asking what was taking so long.

After a reorg in 2023, it was a headless project with no owner and a lot of pressure.

I went to my manager and said, "put me in coach."

First: figure out what we actually have

With an entirely new team of product, design, and engineering, the first thing we did was assess the situation. Outside a massive Figma file and a partially coded frontend, almost nothing was defined or documented. Nobody could tell us what done looked like.

So we started with the Figma file.

I reviewed the entire file and started cutting anything that was clearly nice-to-have. I met with the VP of Tech & Innovation to understand his expectations. I met with engineering to get clarity on what was already coded, what was left, and what was actually feasible.

The modular checkout had to go

Someone had introduced the idea that the checkout steps could be rearranged at will. No clear rationale, massive complexity. It was exactly the kind of thing that sounds interesting in a roadmap meeting and becomes a nightmare in production.

It went.

Then we built it properly

My design partner and I started fresh: new Figma file, full checkout spec'd from scratch using our newly established design spec standards. Reduced scope. Clear documentation. An actual source of truth in just one week.

With that in hand, product and engineering could finally define what work was left and estimate it accurately.

They estimated nine months. It shipped in six.

What I'd take with me

When a project has been spinning for two years, the problem is almost never the design. It's that nobody has named what done actually means. Once we had that, everything else followed.