The Complete Guide to Startup Sourcing

Startup sourcing is the active search for companies worth evaluating before any of them apply. Where screening filters the deals you already have, sourcing decides which deals you get in the first place. The quality of everything downstream is capped by the quality of what you sourced.

Most funds and programs treat sourcing as something that happens to them. Applications arrive, warm intros land, a partner forwards a deck. That is inbound, and inbound has a bias: you only see what already knows you exist. The best company for your thesis this quarter may never send you anything. Sourcing is how you go find it anyway.

This guide covers what sourcing is, why inbound alone is not enough, how to source against a thesis instead of a keyword, where startups leave public traces, a step-by-step workflow you can run this week, how to qualify before you spend a single conversation, and the mistakes that quietly waste a sourcing effort.

In this guide

• Inbound is not a strategy, it is a filter
• Source against a thesis, not a keyword
• Where startups leave public traces
• A practical sourcing workflow
• Signals that separate momentum from theatre
• Qualify before you reach out
• From watch to pipeline, without changing tools
• Common mistakes
• FAQ

Inbound is not a strategy, it is a filter

Inbound deal flow feels like abundance. A well-known accelerator can receive thousands of applications a cycle. But volume is not coverage. Inbound over-represents companies that are good at applying, plugged into the right networks, and already looking for you. It under-represents the quiet operator two cities over who is heads-down building and has never heard your name.

Relying on inbound means your pipeline is shaped by who found you, not by who fits you. Outbound sourcing flips that. You start from your thesis and work outward to the companies that match it, whether or not they were ever going to reach out.

The real trade-off

Inbound is cheap and biased. Outbound is more work and less biased. Neither is a full strategy alone:

• Inbound fills the top of the funnel at almost no cost, but its composition is decided by your reputation and reach, not your thesis. If your brand skews toward one geography or sector, so will your inbound, silently.
• Outbound lets you correct for that bias on purpose. It is how you reach the companies your reputation does not, and it is the only lever that improves coverage rather than just volume.

A serious pipeline runs both, and treats inbound as one input to be qualified like any other, not as a privileged one because it arrived unsolicited.

Source against a thesis, not a keyword

The common failure in outbound sourcing is searching by keyword. "AI healthcare seed" returns ten thousand results and no judgment. A keyword tells you what a company says about itself. It does not tell you whether the company fits what you actually back.

Sourcing against a thesis means encoding your real criteria first: the stage you invest at, the geographies you cover, the sectors you understand, the business models you believe in, and the lines you will not cross. That written standard is what turns a raw list into a qualified one. (For how to write that standard, see the investment thesis and anti-thesis entries in the glossary.)

What a usable thesis looks like

A thesis vague enough to fit everything filters nothing. A usable one is specific on at least four axes:

• Stage. Pre-seed, seed, Series A. This alone removes most of the noise.
• Geography. The markets you can actually support and evaluate.
• Sector and model. Not "tech" but "vertical SaaS with a usage-based model" or "hardware with a regulated go-to-market."
• The anti-thesis. The things you will not back no matter how good they look. This is the axis most funds skip, and it is the one that saves you from the deal that scores well on paper and violates a line you drew on purpose.

A thesis-driven source list is smaller and far more useful than a keyword dump. Fifty companies that genuinely match your stage, geography and model are worth more than five thousand that merely mention your sector.

Where startups leave public traces

You do not need private data to find and pre-qualify most companies. Early-stage companies leave a surprising trail in public and official sources:

• Company registers. Incorporation, standing, officers and filings sit in official registries: SIRENE and INPI in France, SEC EDGAR and Secretary of State records in the United States, Companies House in the UK, GLEIF across the EU. This is the ground truth for whether a company is real, how old it is, and who runs it.
• Funding signals. Public raise filings (such as SEC Form D in the US, Form C for equity crowdfunding) and press coverage mark who raised, roughly when, and often how much.
• Patents and IP. USPTO, EPO and INPI show what a company has actually filed, which is a harder signal than a slide claiming proprietary technology.
• Hiring and product. Job postings reveal roadmap and momentum before any announcement. A company hiring three sales roles is telling you something a press release will only say in six months.
• Indirect signals. When the real numbers are private, proxies stand in: headcount growth, web traffic trends, community footprint, review volume. None is proof on its own, but together they separate momentum from a landing page.

The point is not to trust any single source. It is to triangulate. A company that shows up across registers, filings and hiring signals is a different prospect from a name that appears only in its own marketing.

A practical sourcing workflow

Sourcing is not magic, it is a repeatable loop. One version that works:

1. Write the query, not the keyword. Turn your thesis into a search you can run: stage, geography, sector, and a signal (recently raised, hiring, filed a patent). "Seed-stage French climate-hardware companies that posted engineering roles in the last quarter" is a query. "Climate tech" is a keyword.
2. Cast across sources. Run it against registers, funding filings, hiring boards and the web. Each source sees a different slice; no single one is complete.
3. Deduplicate and name the entity. The same company appears under slightly different names across sources. Resolve it to one real, registered entity so you are not chasing ghosts.
4. Pull the public trace. For each candidate, gather what the public record already shows: founding, standing, funding, patents, headcount trend.
5. Pre-qualify against the thesis. Score fit before contact (see below). Anything that fails does not enter the pipeline.
6. Verify the load-bearing claims. For the survivors, check the two or three claims your interest rests on before the first call.

The discipline is that qualification and a first verification happen before outreach, not after. That is what protects your scarcest resource, attention.

Signals that separate momentum from theatre

Early-stage companies are good at looking busy. A few signals are harder to fake than a landing page:

• Hiring velocity in the right roles. Engineering and sales hires signal real traction. A wall of "growth hacker" posts signals a burn.
• Repeat, dated public filings. A company that has filed across multiple periods has a paper trail. A single recent incorporation with big claims deserves more scrutiny.
• Third-party mentions with substance. A customer naming them, an analyst citing them, a registry confirming them. Self-published claims are the starting point, not the evidence.
• Consistency across sources. The strongest signal is boring: the same story holds up in the register, the filing, the hiring board and the press. Inconsistency is where you dig.

Qualify before you reach out

The expensive resource in sourcing is not the list. It is attention. Reaching out to a poorly matched company costs a conversation, a follow-up, and the opportunity you did not spend on a better fit. Upfront qualification is what protects that attention.

Qualification means running each sourced company against your thesis before it enters your pipeline, not after. Does the stage match? The geography? The model? Does anything about it hit your anti-thesis? A company that fails those checks should never reach your calendar, no matter how interesting the one-liner.

This is also where verification starts to matter. A sourced company often comes with claims attached, from its site, its press, its filings. Checking the material ones against public sources before the first call means you walk in already knowing what holds up. (This is the same discipline that powers screening. See The Deckwise Method and Manual vs AI Screening.)

From watch to pipeline, without changing tools

Sourcing is not a one-time list. Your thesis stays stable while the market moves, so the useful posture is continuous: a standing watch that surfaces new companies as they cross your criteria, feeding a pipeline you actually track.

The break most teams hit is tooling. Sourcing happens in a browser and a spreadsheet, qualification happens somewhere else, and the pipeline lives in a third place. Every handoff loses context and slows the loop. Keeping sourcing, qualification and pipeline in one place is what turns occasional prospecting into a habit that compounds. (See Deckwise vs Spreadsheets for why the handoff is where most of the cost hides.)

Common mistakes

• Sourcing by keyword instead of thesis. Produces volume, not fit.
• Trusting a single source. One mention is a lead, not a fact. Triangulate.
• Skipping qualification. An unqualified list just moves the filtering cost to your calendar.
• Treating sourcing as a one-off. The market refreshes. A stale list is a missed quarter.
• Confusing a good deck with a good company. Sourcing surfaces candidates. It does not excuse you from checking them.
• Reaching out before verifying the load-bearing claim. The first call is worth more when you already know which claims hold.

Frequently asked questions

What is the difference between sourcing and screening? Sourcing finds companies worth evaluating before they apply. Screening filters and ranks the ones you already have. Sourcing feeds the top of the funnel; screening decides what makes the shortlist.

Do I need paid data to source well? No. Most companies can be found and pre-qualified from public and official sources: registers, funding filings, patents, hiring boards. Paid data can add coverage, but the public trace is enough to start and often enough to decide.

How many companies should a sourcing pass produce? Fewer than you expect, and that is the point. A tight thesis produces a short, high-fit list. If a pass returns thousands, it was a keyword search, not a thesis search.

When does outbound sourcing beat inbound? Whenever your reputation does not reach the companies that fit you, which is most of the time for a focused thesis. Inbound covers who knows you; outbound covers who fits you.

The bottom line

Sourcing sets the ceiling on everything downstream. Source by thesis, not by keyword. Triangulate the public trace instead of trusting a single mention. Qualify and verify the load-bearing claims before you spend a conversation. And keep the loop continuous, in one place, so it compounds instead of restarting every cycle.

How this shows up in Deckwise

Deckwise treats sourcing as the first of three pillars. The Prospecting pillar sources companies against your thesis and geography from public sources, qualifies them upfront, and feeds a pipeline in the same place you screen and verify. The signals it reads are the public ones above, triangulated, not a single feed taken on faith. The benefit is not "more leads." It is that the company you should have found does not stay invisible because it never applied.

Next: The Complete Guide to Startup Screening — turning a sourced pipeline into a defensible shortlist.