A signal is a specific, checkable piece of evidence about a person or company, such as a shipped product, a revenue figure, a patent, a regulatory filing or a code contribution, that a reviewer or model uses as an input to a score because it says something about likely future performance.
What makes a strong signal
Not every fact about a founder or candidate is a signal. A fact earns the name when it changes what you expect to happen. Strong signals tend to share five properties:
- Specific. "Grew a product from zero to 40,000 weekly users" rather than "growth-minded".
- Checkable. Someone else can confirm it from a source: a filing, a repository, a reference, a dashboard.
- Close to the outcome. Selling to the target customer says more about a sales hire than the name of their university.
- Costly to fake. Economists call these costly signals: they are credible because a weaker candidate could not easily produce them (Connelly et al., 2025).
- Recent. A launch last quarter counts for more than one from eight years ago.
Everything else is noise: details that feel relevant but do not move the odds. The signals that pass these tests become the inputs to a scoring model.
Examples of signals in VC, PE and hiring
| Decision | Strong signals | Weak signals or proxies |
|---|---|---|
| Seed investment | Founder has built in this market before; paying customers; fast shipping cadence in the public repository | Accelerator brand; follower count; deck design |
| PE target screen | Revenue and margin trend from filings; customer concentration; leadership changes | Press mentions; award lists |
| Engineering hire | Code they wrote that others rely on; systems they ran in production; structured interview answers | Employer brand; years of experience on their own; school |
| Go-to-market | Repeat purchases; expansion revenue; buyer research activity (intent data) | Website visits from unknown sources |
A proxy is a signal standing in for something you cannot see directly. Employer brand is a proxy for skill. Proxies are not useless, but they are where bias tends to enter, because access to the proxy is not evenly spread. Checking which proxies a score leans on is a large part of bias mitigation.
Timing signals work differently. A new funding round or a new senior hire says little about quality but a lot about when to reach out, which is what cold outreach scoring uses them for.
What the research says about which signals predict
In hiring, the most thorough recent review of selection methods, by Sackett and colleagues, concluded that the validity of many methods had been overestimated in older work, and that structured interviews came out as the top-ranked procedure (Sackett et al., 2022). The signal in a structured interview is the candidate's answer to the same question everyone else got, scored on the same scale.
In venture capital, a survey of 885 VCs at 681 firms found that when choosing investments, VCs see the management team as more important than business characteristics such as product or technology (Gompers et al., 2020). That is why founder scoring puts so much effort into finding checkable evidence about the team, rather than relying on impressions from one meeting. Signals also drive VC deal sourcing: a founder leaving a job, or a repository gaining users, can surface a company before it raises.
Worked example: separating signal from noise
A partner's notes on a fictional founder, Priya at Quarry Labs, list six facts. Sorted:
- Signal: Led the data platform team at a logistics company for four years (checkable via references, close to the market Quarry sells into).
- Signal: Three paying pilots signed in the last 90 days (checkable via contracts, recent).
- Signal: Public repository shows weekly releases since launch (checkable, costly to fake).
- Proxy: Studied at a well-known university (weakly related to the outcome).
- Noise: "Very confident presenter".
- Noise: 12,000 social media followers.
Scoring on the first three gives a clear, arguable picture. Letting the last three in adds variance without adding information.
Common mistakes
- Counting the same signal twice. "Raised a large seed round" and "backed by a top fund" are often the same fact.
- Mixing units. Revenue, headcount and ratings need normalization before they can be combined.
- Trusting signals nobody checked. Self-reported numbers are claims until they are confirmed. Verification is the point of evidence-based scoring.
- Stale data. A headcount from 18 months ago may describe a different company. Refresh it through data enrichment.
How ScoringFactory approaches signals
ScoringFactory learns which evidence a team's past yes and no decisions actually turned on, applies that to every founder, company or candidate in the pool, and cites each score to the line in the record it came from. The team sees the receipts and makes the call. See the flow.
Frequently asked questions
What counts as a signal in deal sourcing or hiring?
Any specific, verifiable fact that changes your expectation of success. In deal sourcing that might be a founder's prior experience in the market, paying customers or a fast release cadence. In hiring it might be work samples, systems the candidate ran in production, or answers to structured interview questions scored against a rubric.
What is the difference between signal and noise?
Signal changes the odds of a good outcome; noise does not, even if it feels relevant. A founder's three signed pilots are signal. Their presentation polish usually is not. Separating the two is hard in a meeting, which is why writing down the evidence behind each score helps: noise is hard to cite.
Which signals predict founder or candidate success?
For candidates, research favours structured interviews and work samples over unstructured conversation and credentials. For founders, VCs report that the team matters most, but which team facts predict success varies by fund and stage. The reliable approach is to test your own signals against your own past outcomes.
Is a signal the same as a data point?
No. A data point is any recorded fact. A signal is a data point that has been shown, or is reasonably believed, to relate to the outcome you care about. A company's founding year is a data point. Its revenue growth over the last four quarters is usually a signal.
Sources
- Sackett et al. (2022), Revisiting meta-analytic estimates of validity in personnel selection, Journal of Applied Psychology (APA)
- Connelly, Certo, Reutzel, DesJardine and Zhou (2025), Signaling theory: state of the theory and its future, Journal of Management (SAGE)
- Gompers, Gornall, Kaplan and Strebulaev (2020), How do venture capitalists make decisions?, Journal of Financial Economics