Early Startup Signals: A Field Guide for VC Sourcing Teams

Early startup signals are public traces of a company being built before it raises: a new legal entity, a hiring plan, a domain, code, a founder changing jobs. Each one is weak on its own. The useful work is knowing the base rate behind a signal, the false positives it usually produces, how quickly it goes stale, and which second signal confirms it. This guide covers all four, signal by signal, and ends with a log format a sourcing team can adopt this week.

If you want the broader workflow for reaching founders before a round opens, start with our post on how VC funds find startups before they fundraise. This post goes one level deeper, into how to read each signal.

Start with the base rate: how few new businesses become venture companies

Most signals start from a very large pool, and only a tiny part of that pool will ever be a venture-scale company. U.S. federal data shows how fast the pool narrows.

The Census Bureau's Business Formation Statistics track applications for an Employer Identification Number (EIN). For September 2026 (BFS release, published 14 October 2026, seasonally adjusted):

Shares are our arithmetic on the Census figures. The last row is a Census projection of how many September applications will become businesses with payroll within two years, not a count of businesses that already exist.

The Census definitions page explains what moves an application into the "high-propensity" group: it comes from a corporate entity, says it is hiring, gives a first date for paying wages, or sits in certain industries. Those criteria are worth copying into your own filter, because they are the same questions a sourcing team asks: is this an operating company that plans to employ people?

The next step narrows much further. Manju Puri and Rebecca Zarutskie followed U.S. employer firms from birth using the Census Bureau's Longitudinal Business Database. They found that venture-financed firms averaged about 0.1% of all new firms created over 1981 to 2001, rising to 0.2% in the late 1990s (NBER Working Paper 14250, Section III; published in the Journal of Finance, 2012). That data is old and the venture market is much larger now, so treat the ratio as an order of magnitude, not a current rate.

The practical point: a single formation signal, taken alone, will be wrong about venture fit almost every time. A signal is a reason to look, never a reason to reach out.

The signal timing ladder

Signals tend to appear in a rough order. Where a signal sits on this ladder tells you how early you are, and how much time is left before the round.

1. Founder intent. A departure from a role, a change of title to "building something new", a burst of public writing on one problem.
2. Legal and administrative setup. Incorporation, an EIN application, a registered trademark or domain.
3. Building. A product site, a waitlist, code, design work, early documentation.
4. Team growth. First hiring posts, especially for roles beyond the founders.
5. Capital already committed. A Form D filing with the SEC.

The last rung matters because teams often mistake it for an early signal. Under Regulation D, a company files Form D within 15 days after the first sale of securities, and the SEC defines that date as the day the first investor is irrevocably committed to invest (SEC, Filing a Form D notice). By the time a Form D appears, at least one investor has already committed. It is useful for mapping a market and for timing follow-on conversations. It is not a pre-round signal.

Field notes on each signal

Each entry below gives what the signal can tell you, the false positives it usually produces, and what to pair it with. The freshness windows are our working defaults for an active sourcing radar, not research findings; tune them to your stage and sector.

Founder departures

A departure is the earliest signal and the least certain. Its value depends almost entirely on context: who left, from where, with whom, and what they did before. Two people leaving the same team in the same month is a far stronger signal than one. Check whether the person has started an entity or registered a name within the window; if not, park the profile and let the next signal bring it back.

Formation and registration records

Formation records are broad by design. The Census numbers above show that roughly seven in ten U.S. business applications in September 2026 did not meet even the high-propensity criteria. Filter formation records with the same questions Census uses: is it a corporation, does it plan to hire, does it plan to pay wages, and does its industry fit your thesis. A Delaware C corporation with two named technical founders is a different signal from a single-member consulting LLC.

Domains and product sites

A domain on its own says little. A live page with a product description, a waitlist or named founders says much more. Track the change, not the registration: the step from parked domain to live product page is often the more useful event.

Public code activity

Public code is a partial view of what is being built. GitHub's Octoverse 2025 report counted more than 121 million new repositories created in its September 2024 to August 2025 year, so a new repository alone is noise. The same report found that 81.5% of contributions happened in private repositories. Many serious teams build in private, so the absence of public code is not evidence of absence. Weight commit regularity, the number of contributors and whether the code connects to a product.

Hiring posts

Hiring is the strongest of the pre-round signals because it costs money and commits the founders to a plan. It is also the latest. Look at what is being hired: a first engineer or designer at a two-person company says more than a long list of sales roles, which may indicate a company already past your stage.

Founder writing and talks

Public writing shows how a founder thinks and where their attention is going. Judge it over time. A founder who has written about the same problem for a year, then goes quiet and registers a company, is a better lead than one viral post.

How to combine signals without fooling yourself

Combination is where signals become useful, and also where teams over-count. Three rules keep the picture honest:

• Count independent evidence, not repeated evidence. A company announcing its launch in five places is one signal, not five.
• Require one signal of commitment. Departures, domains and writing are cheap. Entity formation with hiring plans, multiple contributors or a live product cost something. Ask for at least one costly signal before a company reaches an analyst.
• Keep the reasons visible. A single score hides why a company surfaced. Record which signals fired, when and from where, so the next person can check the work.

A signal log your team can use

Log each company with the same fields. It makes reviews faster and stops two people researching the same company twice.

Review the radar on a fixed cadence. Anything past its freshness window with no new signal should be archived with a note, not left to clutter the list.

Where software helps

Collecting signals, checking them against a thesis and keeping a radar current is repetitive work that software handles well. Deciding what counts as a costly signal for your fund, and which founders to back, stays with your team.

Juno is Sonexa Labs' ready-to-use sourcing product. It helps venture investors discover emerging companies through early building activity, investigate founder and company context, keep companies on a radar and track outreach.

Funds that want sourcing built around their own mandate, signals and review process can commission a fund-specific discovery system from our Tailored Solutions team instead.