Adverse Selection: The Hidden Cost of Warm-Intro-Only Dealflow
Every investor who relies on warm introductions is running a filter whose selection criterion is proximity to their existing network. This is individually rational and collectively expensive, and the cost is invisible by construction, because the companies it excludes never appear in any report you will ever read.
This is the economics of that trade, honestly stated, including the parts that favour the warm intro.
Key takeaways
- The warm intro is a high-precision, low-recall filter. It is right about what it surfaces and silent about what it misses.
- The cost is type II error - good companies never evaluated - and it is untracked by every investor, everywhere.
- The filter correlates with geography, education and background, so its misses are systematic rather than random.
- Concentration of capital in a few geographies means the excluded set is large and growing.
- The fix is not "read everything". It is making evaluation cheap enough that reading widely is affordable.
The trade being made
A warm intro compresses a twenty-minute screening cost to roughly zero by outsourcing triage to a trusted third party. That is a real efficiency gain and nobody should pretend otherwise.
What is being traded away is recall. The filter only surfaces companies within one or two hops of your existing network. Everything beyond that is not rejected; it is never seen.
| Precision | Recall | |
|---|---|---|
| Warm intro only | High | Very low |
| Open inbound, unstructured | Low | High |
| Open inbound, structured and scored | Moderate to high | High |
The third row is the interesting one and it did not exist as an option until recently. Most of the industry's instincts were formed when the only choice was between the first two.
The invisible cost
An investor can tell you their hit rate on investments made. No investor can tell you their hit rate on companies they never evaluated.
The asymmetry is total:
- A bad investment is visible, memorable, discussed at length, and written into the fund's history.
- A missed company produces no artefact at all. There is no record, no feedback, no learning event. It simply never happened.
This is why the warm-intro filter persists despite being widely understood to be lossy. There is no feedback loop that would ever generate pressure to change it. The errors it makes are precisely the errors that never report themselves.
The only partial defence is to track pass reasons and revisit them. If you log a company as "too early" and check back eighteen months later, you occasionally discover it raised a strong Series A from someone else. That is a data point about your process, and it is the only kind available.
Why the misses are systematic
If network proximity were randomly distributed, warm-intro filtering would cost recall but not introduce bias. It is not randomly distributed.
Network proximity to venture capital correlates strongly with: living in a small number of metropolitan areas, having attended a small number of universities, having worked at a small number of companies, and being demographically similar to existing capital holders.
So the excluded set is not a random sample of companies. It is concentrated in specific founder profiles and specific geographies, which means the filter produces a portfolio that is systematically narrow along dimensions that have no established relationship to returns.
The global data makes this concrete. Per Crunchbase News data, the US share of global venture fell from roughly 83% in Q1 2026 to about two thirds in Q2, as Europe posted its strongest quarter in four years and Asian markets accelerated sharply. Capital formation is globalising faster than investor networks are. An investor whose dealflow depends on a network built in one city is sampling from a shrinking share of the opportunity set, and the gap widens every quarter.
The arithmetic of a wider funnel
Take an investor doing six deals a year from 400 sourced companies, nearly all warm.
Suppose an open channel adds 600 companies a year, of which the true quality distribution is somewhat worse - say the hit rate is half that of the warm channel. That still adds roughly 300 quality-equivalent companies to a funnel of 400.
The binding constraint is not whether those companies are worth seeing. It is whether they can be screened. At twenty minutes each, 600 companies is 200 hours a year, which is unaffordable and will be abandoned by March. At ninety seconds each, it is fifteen hours, which is trivial.
The whole argument reduces to screening cost. Investors do not avoid open inbound because they believe it is worthless. They avoid it because unstructured inbound is unaffordable to read, and it is unstructured because there is no standard for how a company presents itself.
Fix the format and the economics of the filter change completely. That is the operating detail in pre-screening at scale.
What a wider funnel requires
Three things, in order:
- A structured intake. Companies arriving in a comparable field set, not as PDFs. Without this, nothing else works.
- A written thesis filter. Published, so founders self-select, and applied identically so gate one costs thirty seconds.
- A consistent first-pass score. So that a company arriving with no social proof gets the same reading as one arriving through a partner.
Point three is the one that actually addresses the bias. A score computed from a standard field set does not know who introduced the company. That is its entire value.
The honest counter-argument
The strongest case for warm intros is not screening cost. It is that the introducer's judgement carries information that no field set captures: how the founder behaves under stress, whether they are honest when it is inconvenient, whether people who worked with them would do it again.
That is real, it is valuable, and structured data does not replace it.
But it is an argument for warm intros as a signal, not as a gate. A referral should raise a company's priority. It should not be the mechanism by which a company becomes visible at all. Those are different roles and conflating them is precisely where the cost accumulates.
Frequently Asked Questions
Usually it refers to the risk that the deals available to you are the ones better-informed investors passed on. The related and larger problem is selection by network proximity, where the filter systematically excludes companies outside your existing connections.
Yes, if they can screen it cheaply. The cost is reading time, so the answer depends entirely on the format companies arrive in.
They predict that someone the investor trusts thought it worth their social capital. That correlates with quality and correlates more strongly with proximity.
Structured intake, a published thesis filter, and a consistent first-pass score. Screening cost is the constraint, not quality.
Read next
The Modern Dealflow Stack: From Inbox Chaos to a Scored Pipeline
Pre-Screening at Scale: Scoring Startups Before the First Call
You Cannot Compare Two Startups Today. Here Is Why That Is a Data Problem
Warm Intros Are a Filter You Did Not ChooseRaiseTalks matches startups to investors on sector, stage, ticket size and geography, and founders can push their data room only once it clears a readiness threshold. Structured intake, no introduction required. Start your 30-day trial.