M&A Marketplace
Match.Asia Intelligent Matching Engine
Results
What we moved.
2,000+
Duplicate CRM records resolved in the foundation cleanup
30+
Custom reports across three leadership dashboards (buyers, sellers, pipeline)
4 weeks
From Phase 2 kickoff to a working agentic matching engine
< 1 hour
Match-list creation target; the manual process took two to three days
Challenge
Match.Asia connects companies that want to sell or raise capital with the right investors across Asia. The matching is the business. And it was entirely manual: building a buyer list for one mandate took an analyst two to three days, stitched together from ChatGPT research, contact-database lookups, and LinkedIn checks, one investor at a time.
The CRM underneath could not support anything smarter. It carried more than 2,000 duplicate records. Investment criteria sat in free-text fields that no filter could read, and nobody could see pipeline health across buyers and sellers in one place. Before AI could rank matches, the data had to become trustworthy.
Solution
We built the platform in two phases: first make the data solid, then make it think.
The foundation. A structured hard-criteria framework (sector, ticket size, geography, deal type, valuation) replaced free-text guesswork on buyer and seller records. We resolved the 2,000+ duplicates and standardised formatting with automated data-quality rules instead of manual scripts. Three leadership dashboards with more than thirty custom reports gave the team live visibility over buyer onboarding, seller mandates, and pipeline. Branded onboarding forms fed clean data in from day one.
The matching engine. When a new seller mandate is listed, the engine scores every buyer in the CRM against it: sector alignment (40%), ticket size (30%), geography (20%), investment style and recency (10%). Each match lands on a review board in HubSpot with a written rationale explaining the score. The team reviews reasoning, not black-box numbers. For investors outside the CRM, a discovery workflow finds lookalike buyers on the open web and enriches them in three layers: firmographics first, then verified contact details, then AI web research. That replaced the manual research chain. Approved matches flow into personalised multi-step outreach sequences.
Before committing to the build, we tested the hardest assumption with a live scoring run against a varied investor panel, including one deliberately poor fit as a control. The control scored 12 out of 100. Every score came back with source-grounded reasoning specific enough to catch mandate mismatches that a keyword filter would miss.
12 / 100
Engine output, validation run, control buyer
"The buyer is a shipping and logistics group with no identifiable investment mandate, M&A track record, or active investment presence in financial services or fintech. The enrichment data shows zero confidence, and there is no evidence of strategic interest in SE Asian SME lending platforms or fintech investments."
The stack. HubSpot Customer Platform Pro and Data Hub Pro as the system of record. The i10X agent platform runs the matching and discovery agents. Enrichment layers through Breeze Intelligence, Surfe, and AI web research. Reporting stays native in HubSpot, on the dashboards the leadership team already uses.
The engagement continues as a monthly retainer: sprints that refine scoring weights, outreach sequences, and reporting as the marketplace grows.
Services used