The $1M project I killed on purpose.

Everyone wanted this one to work.

A healthcare client asked us to scope an AI system that would aggregate patient-level data across provider networks to surface care-gap opportunities. Leadership was bullish. Engineering had a clear architecture. The pilot would have hit $1M in commitments before the first integration sprint.

My job in discovery was to de-risk, not to cheerlead.

Two weeks in, I found the math didn't work.

The data sources the system needed contained PHI and PII that couldn't leave the originating networks without BAA renegotiation — a 6-to-12 month legal path with uncertain outcome. Even with data in place, the ROI model broke: projected uplift on care-gap closure priced against operating cost fell below the break-even line in every scenario I could build.

Kill it. Write down why. Move the team.

I wrote a one-page kill memo: compliance blockers, revised ROI model, alternatives considered. I presented it to the steering group expecting pushback. I got questions — good ones — and then a decision to terminate. The team rotated onto a parallel initiative that shipped three months later.

This is the case study I'm most proud of. Anyone can ship features. Not everyone will recommend against shipping when the org has already imagined the launch.

What this taught me about discovery.

Compliance isn't a Phase 2 problem. In regulated domains, the legal path determines the critical path. Validate BAAs, DPAs, and data-residency before you validate the feature.

ROI modeling is a PM deliverable. If you can't draw the unit economics in discovery, you can't defend them in executive review. Learn the spreadsheet.

The kill memo is a PM artifact. A crisp, written recommendation against shipping is one of the hardest and most valuable things a PM produces. Most don't. The ones who do become the ones leadership trusts.