Building a Sourcing Engine That Compounds

Deal flow feels like luck when you're new and looks like infrastructure when you study firms that sustain it. The best sourcing engines are boring: defined channels, assigned owners, logged touches, measured conversion. Here is how to build one that gets stronger every quarter instead of resetting to zero.

Channel 1 — The referral network

The highest-conviction deals arrive through people: founders you've backed, angels who trust your speed, co-investors who want you in their rounds, and operators who hear about companies before anyone else. This channel compounds only if you service it:

• Founders first. Portfolio founders who feel genuinely helped refer their best friends. A reputation for fast, honest passes matters almost as much — founders talk about how you say no.
• Give allocation to get allocation. Syndicate leads and angel groups remember who moved quickly and behaved well in tight rounds.
• Map it. A relationship-intelligence CRM like Affinity or Attio, auto-capturing email and calendar activity, tells you who on the team actually knows the person who knows the founder — before you need the intro.

Channel 2 — Thesis-driven outbound

Waiting for rounds to form means paying formed-round prices. Outbound gets you into companies before there's a process:

• Write theses narrow enough to name the interesting companies in a space. "AI" is not a thesis; a specific workflow, buyer, and why-now is.
• Use signal platforms — Harmonic, Specter, Crustdata and peers — to detect companies at formation: founding-team departures from notable employers, stealth incorporations, early hiring, repo and traffic momentum.
• Reach out like an investor, not a lead-gen bot. One specific observation about their business beats any sequence template. The goal of the first touch is the second touch; most founders you contact won't be raising, which is precisely the point.

Channel 3 — Brand and inbound

Content, a sharp point of view, and visible wins turn your firm into a destination. This is the slowest channel to build and the cheapest to run at scale: one genuinely useful essay in your thesis area outperforms months of cold outreach because it makes the right founders come to you, pre-sold. Publish what you actually believe, show your work, and let your portfolio's founders vouch publicly. Inbound quality, not volume, is the metric.

Channel 4 — Institutional nodes

Accelerator demo days, pre-seed funds that feed your stage, university programs, and angel syndicates are structured aggregation points of early companies. Cover them systematically: know the partners, show up between batches rather than only at demo day, and track which nodes have historically produced companies that fit your thesis. Not all nodes are equal for your strategy — measure graduation into your portfolio, then concentrate.

The system underneath: log everything

A sourcing engine without a system of record is a group chat. Non-negotiables:

• Every company seen gets a record, an owner, a source, and a stage — including passes.
• Write down why you passed. The anti-portfolio review (below) is impossible without it.
• Pipeline review weekly, with the same seriousness as an IC meeting. Sourcing dies as a side task.

Measure the engine

Three numbers tell you most of the truth:

1. Coverage — of the relevant rounds announced in your thesis areas last quarter, what share did you see before close? This is the single best health metric for a sourcing engine, and most firms can't answer it.
2. Win rate — of the deals you tried to do, how many did you get? Low win rate is a brand, speed, or terms problem, not a sourcing problem.
3. Channel yield — which channel produced the deals you actually did, per hour invested? Fund the channels that produce; kill the ones that only produce meetings.

Close the loop

Twice a year, run the anti-portfolio review: every pass that went on to raise a strong up-round, alongside the reason you logged at the time. Patterns emerge fast — a price discipline that was really a stage mismatch, a founder archetype you systematically underrate. This review is uncomfortable and is also the only mechanism by which a sourcing engine truly compounds, because it upgrades the judgment sitting on top of the pipeline.

Build the channels. Log the flow. Measure coverage. Review the misses. Do it for eight quarters and your deal flow will look like luck to everyone else.