Every inbound founder at Harlem Capital gets rated on the same scorecard. Not because founders are interchangeable, but because a partner's judgment is the most expensive and least consistent thing in the fund. Here is exactly what we rate, why we built it this way, and how every score ties back to something you can point to.
I have sat through roughly two thousand founder pitches over the last decade. For the first few hundred, a free-form memo worked fine. I knew what I meant when I wrote "strong team" and so did the two other partners reading it. By deal five hundred, the memo had quietly drifted. "Strong team" from me in January did not mean the same thing as "strong team" from me in October, and it definitely did not mean the same thing as when a newer partner wrote it. Nobody noticed until we compared notes on two deals that should have scored the same and didn't.
Why the memo breaks at scale
A prose memo is a great record of a conversation and a terrible unit of comparison. It rewards whoever writes the most persuasively about the founder they liked, which is often the founder who was most charming in the room that day, not the founder with the best odds. A recent rundown of what top-tier funds actually screen for in a 2026 breakdown of forty questions VCs ask before funding makes the same point from the other side of the table: the firms that convert faster are the ones that walk into the first meeting with a fixed set of dimensions, not a blank page. Founders prepare for structure. Investors should bring structure too.
What we built is a formal scorecard, not a vibe check with a nicer font. Every inbound founder, whether they came through a warm intro or cold-emailed me on a Tuesday, gets rated against the same fixed set of dimensions before any partner writes a paragraph of prose. The prose still matters. It just comes after the numbers, not instead of them.
The seven things we rate
The dimensions took three years to settle on, mostly by tracking which early ratings actually predicted a company still being alive and growing eighteen months later. Here is what survived:
- Problem obsession. How long has this founder personally lived inside the pain point, and can they describe it in a customer's words rather than a market-research phrase.
- Distribution instinct. Not "we'll figure out growth," but a specific channel they have already tested, with a number attached to it.
- Operating depth. Have they actually built or run the thing before, separate from whether they can tell a good story about the vision.
- Capital discipline. What did they do with the last dollar they raised, not what they promise to do with the next one.
- Coachability under contact. How they respond in the room when a partner pushes back on their own idea, in real time.
- Gap awareness. Whether they can name what they are bad at and who they are already recruiting to cover it.
- Founder-market fit. Why this founder, this problem, this decade, and not some other founder who is also smart and also hardworking.
A memo tells you how a partner felt about the pitch. A scorecard tells you why, in a form the next partner can check against their own read.
Evidence, not vibes
None of these seven ratings are useful if they are just a number a partner felt like assigning. Each score on our sheet has to cite something: a line from the call transcript, a retention number the founder actually showed us, a repo commit history, a reference call quote. This is the same shift a peer-reviewed multidimensional due diligence framework published this year argues for at the academic level: moving away from a single composite score toward a diagnostic profile where every dimension is anchored to a specific, checkable behavior rather than a gut impression dressed up as a rating.
In practice this means our version of evidence-based scoring forces a habit change during the call itself. Instead of writing "founder seems scrappy," the associate on the call has to write down the actual sentence that made them think that, timestamped. If nobody can point to the evidence, the dimension does not get scored yet. It gets flagged for the next conversation instead of guessed at.
The scorecard has caught us twice giving a founder a high operating-depth score based on a job title alone, before anyone had actually confirmed what they shipped. Both times, one line of the sheet forced the question that would have otherwise gotten smoothed over in a glowing memo.
Same bar, every partner
The real payoff shows up months after the first pitch, when a deal comes back for a second look with a different partner in the room. Without a shared scorecard, that second partner is reading the first partner's prose and unconsciously translating it through their own private definition of "strong." With the scorecard, they are looking at the same seven numbers, each one backed by the same kind of evidence, and they can disagree with a specific line instead of a general feeling. We wrote more on how this stops score drift from creeping back in once a fund scales past a handful of partners, because a scorecard that isn't actively recalibrated will drift right back into the same mess a memo did.
This is also what calibration actually means in practice at a fund: not a one-time training session, but a running check that every partner's 7 out of 10 on distribution instinct still means the same thing it meant last quarter. We re-run a sample of closed deals against the current sheet every few months specifically to catch drift before it costs us a deal we should have made, or worse, one we shouldn't have.
I built the first version of this scorecard in a spreadsheet because I got tired of re-litigating my own gut feel. It is more disciplined now, but the reason for it hasn't changed. A scorecard does not replace judgment. It makes judgment something you can actually defend in the room, six months later, to a partner who wasn't there for the call.