The Modern Dealflow Stack: From Inbox Chaos to a Scored Pipeline
Dealflow management is the set of processes an investor uses to source, triage, evaluate and decide on companies. At early stage, most of it still runs on an inbox, a spreadsheet and memory, which works up to roughly the point it stops working entirely - usually somewhere around forty inbound companies a month.
This is what the stack looks like when it is built deliberately: five layers, what each one owes the next, and where the real losses occur.
Key takeaways
- Dealflow is five distinct jobs, not one. Conflating them is why pipelines feel busy and produce little.
- The expensive failure is not a bad investment. It is a good company you never read properly. That loss is invisible and therefore untracked.
- Triage should be cheap, explicit and written down. If your first filter is "does this look interesting", you have no filter.
- Scoring needs to be comparable across companies, which means it has to run on structured inputs. A score derived from a PDF is an opinion with a number attached.
- The decision layer needs a memo and an audit trail, even for a $50K angel cheque. Especially then.
The five layers
| Layer | Job | Typical failure |
|---|---|---|
| 1. Sourcing | Get the right companies in front of you | Over-reliance on warm intros narrows the funnel to your existing network |
| 2. Triage | Decide in minutes what deserves an hour | Implicit criteria, applied inconsistently, never written down |
| 3. Evaluation | Form a comparable view of the company | Every company arrives in a different format, so nothing compares |
| 4. Diligence | Verify the claims that matter | Series A process applied to pre-seed, or none at all |
| 5. Decision | Make, record and learn from the call | No memo, no audit trail, no way to review the process later |
Each layer's output is the next layer's input. When layer 2 outputs "seems good", layer 3 starts from zero.
Layer 1: Sourcing
Most early-stage investors have one real channel: their network. It is high-precision and low-recall. It surfaces companies that are close to you, which correlates with geography, background and school far more than it correlates with company quality.
A deliberate sourcing stack has at least three channels:
- Network and referral. High trust, narrow. Keep it, do not rely on it.
- Open inbound. A public, structured way to be pitched. High volume, low precision, and it only works if layer 2 is cheap.
- Active search. Filtering a catalogue by sector, stage, geography and ticket. This is the channel most angels and micro-funds never build, and it is the one that most changes the shape of a portfolio.
The economics of over-indexing on channel one are covered in adverse selection.
Layer 2: Triage
Triage is a filter, and a filter needs written criteria. Not a rubric with weights - just four or five binary questions you apply identically to every company:
- Is it in thesis? Sector, stage, geography, ticket size.
- Is there a team that can plausibly build this?
- Is there evidence the problem is real?
- Is the round structured in a way I can participate in?
- Is there a reason to move now?
Two out of five is a pass. Four out of five is an hour of your time. Write the criteria down once and apply them the same way every time, or you are not triaging, you are reacting to whichever email arrived when you had energy.
The hard constraint: triage must be cheap. If working out whether a company is in thesis takes twenty minutes of reading an unstructured deck, you will not do it consistently, and your funnel collapses back to warm intros by default.
This is the single strongest argument for structured company records. In our experience, triage on a standard field set takes a minute or two; triage on a PDF can take twenty. The difference compounds across hundreds of companies a year.
Layer 3: Evaluation
Evaluation is where a comparable view gets formed: team, product, market, deal.
The requirement that most stacks miss is comparability. A view of one company is nearly useless. The job is to place this company against the others you could fund with the same money, which requires that the same dimensions were assessed for all of them, from inputs of the same shape.
That is why scoring has to run on structure. A score derived from whatever the founder happened to include in their deck measures the deck. A score derived from a standard field set measures the company. See pre-screening at scale for the operating detail, and why you cannot compare two startups today for why this is still broken across most of the market.
Layer 4: Diligence
The main error at early stage is proportionality in both directions: a Series A diligence process applied to a $250K pre-seed cheque, or effectively no verification at all.
Proportionate pre-seed diligence is short and specific:
- Corporate existence and good standing
- Cap table reconciles to every SAFE and note
- IP assigned from founders and early contractors, in our experience one of the most common and most expensive gaps
- Founder backgrounds verified, not just read
- Two or three customer or user conversations
- The one metric the thesis depends on, verified at source
That is days, not weeks. Full checklist in proportionate diligence.
Layer 5: Decision
Write the memo, even for a small cheque. Not for governance theatre: for the audit trail.
A memo written before the decision forces the thesis to be stated in a falsifiable way. A folder of memos written over three years is the only mechanism by which an investor actually learns anything about their own judgement. Without it you have outcomes and hindsight, which teach nothing.
The minimum useful memo is one page: what the company does, why now, the four-dimension read, the two risks that would kill it, what has to be true, the terms, and the decision. What belongs in a pre-seed IC memo has the full structure.
The stage-gate model that holds it together
Stages should be defined by what has happened, not by how you feel about the company:
- Sourced
- Triaged in
- First call taken
- Data room shared
- Second call or team meeting
- Diligence started
- Memo written
- Term sheet issued
- Committed
- Closed
Plus two terminal states: passed and on hold. This is an illustrative model; collapse or merge stages to fit your process. The discipline is that a company only advances when the gate condition is objectively met. "Seems promising" is not a stage. Founders run the mirror image of this on their side, with their own stage definitions, described in run your raise as a pipeline.
Where the money actually leaks
Not in the bad investments. Those are visible, memorable and endlessly discussed.
The losses are in layer 2. Every good company that arrives through an unfamiliar channel, in an awkward format, on a week when you were busy, and gets a polite pass without a real reading. You never learn about those. They do not show up in any report. There is no feedback loop that would tell you they happened.
The only defence is a triage layer cheap enough that you apply it consistently even on your worst week. Which is, in the end, an argument about data structure rather than an argument about discipline.
Frequently Asked Questions
The process of sourcing, triaging, evaluating, diligencing and deciding on investment opportunities, plus the systems that make it repeatable.
It varies widely, but a commonly cited shape is hundreds sourced, tens evaluated seriously, a handful diligenced, one funded. The ratio matters less than whether triage is consistent.
More than funds do. Angels have no analyst team, so the cost of an inconsistent process is paid entirely in their own time. See angels and micro-funds.
Stage, source, thesis fit, the four-dimension score, next action and owner, and a link to the company's record. Anything else is optional.
Read next
Pre-Screening at Scale: Scoring Startups Before the First Call
The Investment Memo, Standardised: What Belongs in a Pre-Seed IC Memo
Proportionate Diligence: A Pre-Seed Checklist That Is Not a Series A Checklist
Adverse Selection: The Hidden Cost of Warm-Intro-Only Dealflow
The State of Early-Stage Capital 2026RaiseTalks gives investors an eight-stage dealflow Kanban you can map these gates onto, a filterable catalogue of startups with structured records, and an AI pre-score engine that produces a comparable investment memo in one click once you have data room access. Explore the investor side.