A deal-matching system that replaced a spreadsheet
A small mergers and acquisitions firm matched buyers to businesses for sale using a 16-tab spreadsheet and one partner's memory. I built the system that replaced it, and measured it against the partner's own picks before anyone relied on it.
The Challenge
The firm finds businesses for buyers before those businesses reach the public marketplaces. The partner could genuinely do the matching from memory, which is why nobody had replaced it, but nothing the firm knew was searchable by anyone else. And the partners were clear that showing a serious buyer a near miss costs more credibility than it gains attention, so the matching had to be strict.
What I Built
- A database and a web app the team uses daily, covering deals, buyers, pipeline and duplicate review
- A matching engine that shows a buyer a deal only if it meets every condition that buyer stated
- A load of 132 deals and 80 buyers, rehearsed on a local copy first, which caught three faults before any reached live data
- Automatic tests on every change, and daily backups proven by actually restoring one
- A priced test of a commercial platform that would have cost about $3,300 a month at their deal volume, and a written reason to build instead
- Eight signed confidentiality agreements recovered from an archive my own import had skipped over, which turned a two-week chase into nothing
Results
The first score was 21 percent, and that was a bug in my own scoring script. Getting to 31 of 31 took three rounds, and every remaining failure traced back to the data rather than the matching logic. More about this kind of work: custom AI tools.
Curious what's possible?
Let's talk.
Tell me about your business and what's slowing it down. I'll give you an honest take on what to fix first, whether that's marketing, bookkeeping or a tool built for the job.
Get in touch