July 30, 2026 · Pipeline, Process
Deal flow triage: passing fast without missing winners
By Emily Buckley, founder of SamCIO
Deal flow management in VC gets treated as a database problem, and it isn't. Almost no fund fails because it lost track of a company. Funds fail at the top of the funnel in two specific ways: they spend real time on deals they were always going to pass on, and they pass in four seconds on deals that needed six minutes. Both are triage failures, and a better CRM does not fix either one.
Good triage is about deciding what gets a fast no, what gets a real look, and what gets a fast yes, using a rule you can state out loud. The rest is bookkeeping.
The expensive failure is the slow no
Track where the time actually goes and the pattern is consistent. The costly deals are not the ones that get an immediate pass or the ones that go to diligence. They are the ones that sit in a middle state for three weeks: a call, then a follow-up, then a request for materials, then a slow fade to no reply.
That deal consumed several hours across two people and produced nothing, including for the founder, who would have been better served by a no in week one. Multiply by the number of deals in that state and it is often the largest single expense in the fund's week.
The cause is almost never that the deal was genuinely close. It is that nobody wanted to write the pass. A fund with an explicit triage rule converts most of those into a fifteen-minute decision, and the founders on the other end are, in my experience, straightforwardly grateful for it.
Score against the thesis, not against your mood
The rule has to be written down or it drifts. What most funds converge on is a small number of dimensions, weighted, with a stated threshold.
The default in SamCIO is four pillars: thesis fit weighted at 30, market opportunity at 25, team and execution at 25, and traction and terms at 20, each scored out of ten with a written rationale, rolling up to a weighted overall. The weights are adjustable per fund, which matters, because a fund whose edge is sector expertise and a fund whose edge is founder access should not be weighting the same things.
The specific weights are less important than three properties. The dimensions are stated in advance, so you are not choosing criteria after you know the answer. Each score carries a rationale, so a 6 on team is arguable rather than a mood. And every deal gets the same treatment, so deal 60 is judged like deal 6 rather than like whatever you were worried about that month.
Bands, not rankings
A ranked list of deals is not actionable. A banded one is. SamCIO's default bands are simple: 8 and above is pursue, 5 to 8 is diligence, below 5 is pass.
The value is in what the bands commit you to. A deal below the pass line gets a written no this week, not a slow fade. A deal in the middle band gets a specific next step with an owner, not a vague follow-up. The top band gets partner time immediately, because the real competition at the top of the funnel is speed.
Thresholds also make disagreement productive. When someone wants to advance a 4, the conversation is about what the framework is missing, which is a useful conversation. Without thresholds it is just a preference, and preferences do not compound into a process.
Explicit stages beat a status field
A pipeline needs enough stages to be honest and few enough that nobody has to think about which one applies. In practice that is five or six: inbox, screening, diligence, pursuing, portfolio, and passed as a terminal state rather than a deletion.
Keeping passed deals is the part funds get wrong. A pass is a decision with a rationale, and it is the highest-value record in the pipeline for two reasons. It is what makes a re-look possible when the company raises again eighteen months later at three times the price and you need to know what you thought the first time. And it is the only way to audit your own false negatives, which is the number nobody measures because it requires having written down the reason.
The screen a model can do, and the one it can't
Automation is genuinely useful at the top of the funnel and genuinely dangerous at the bottom of the score range.
Parsing a deck, extracting company, sector, stage, and the headline metrics, and producing a first-pass score against a written thesis is mechanical work that scales without loss. It is also the work that degrades worst when done manually under volume, because attention is scarcest exactly when inbound is heaviest.
What a first-pass score cannot do is make the pass decision on its own. Automated screening is well calibrated on obvious fit and obvious mismatch and poorly calibrated in the middle, which is exactly where the interesting deals live. A fund that auto-rejects everything under a threshold will lose a specific category of company: the one whose deck is bad and whose business is good. That founder exists in volume at pre-seed.
The workable arrangement is that the model sorts and explains, and a person clears the bottom band with a quick human read before the passes go out. That read takes seconds per deal when the reasoning is already written out.
Measure the funnel, not the activity
Deal count is a vanity metric. Three numbers are worth tracking:
Time from first contact to decision, by outcome. If passes take longer than advances, triage is broken.
Conversion by stage. A fund that screens 100 and diligences 40 is not triaging, it is doing diligence as triage, which is the expensive version.
Source quality over time. Which channels produce deals that reach diligence, not which produce volume. Most funds discover one high-volume source that has never produced a single investment.
None of these require new tooling. They require that pass decisions get recorded as decisions instead of as silence.
Where the discipline actually comes from
The mechanism that makes triage stick is not the software, it is that the written thesis exists and the scores refer to it. A fund without a written thesis will score deals against whatever it thinks the thesis is that day, and the framework becomes a formality wrapped around the same instinct.
Get the thesis written first, then score against it, then let the bands force decisions on a schedule. The pipeline gets faster mostly by getting more honest.
If the screening layer is what you are trying to fix, the AI underwriting page covers the scoring and first-pass read. For the framework the scores refer to, a thesis-driven investing framework comes first in the order of operations.