Screening

Stage-weighted deal screening, explained

A screen that scores a pre-seed and a Series A on the same rubric will mislead you. Pre-seed has no traction to weigh; Series A has too much to ignore. A stage-weighted scorecard fixes that by shifting the weights as the company matures — so the number means the same thing whether you're screening a napkin or a growth round. Here's exactly how the weights move, and a worked example you can check by hand.

Rubric re-derived from public, widely-documented early-stage evaluation frameworks. You supply every rating; the tool only makes your judgment explicit and comparable.

The five dimensions

Every institutional screen touches the same five things. They map one-to-one onto the sections of the IC memo:

• Team — direct, hard-won insight into this problem; shipped evidence; can hire ahead of the curve.
• Market — bottom-up TAM from buyers × price, a credible "why now," and a wedge that expands.
• Product & Moat — in-market and better on an axis buyers pay for; something that compounds.
• Traction — metrics pulled from systems, a clean time series, retention that holds.
• Deal Terms & Fit — valuation defensible vs. comps; the check fits your fund's ownership and construction.

You rate each 1–5 against fixed anchors (what a "5" looks like vs. a "1"), so a 4 on one deal means the same as a 4 on the next.

Why the weights shift by stage

The weighting changes because what a screen can actually observe changes as a company matures. At pre-seed there's no traction to weigh, so Team and Market carry almost all the signal. By Series A there's real usage data, so Traction earns the most weight and Terms start to matter. Here's the exact table the engine uses (each row sums to 100):

Notice Traction: 5 → 15 → 28. At pre-seed, punishing a founder for traction they can't have yet is noise; at Series A, forgiving thin traction is how you fund a treadmill. These exact numbers are one defensible calibration — the point is that they're written down and you can change them.

The formula

The score is a simple weighted average, rescaled to 0–100:

score = Σ over dimensions of (rating ÷ 5) × weight

Because the weights sum to 100 and each rating ÷ 5 is between 0 and 1, the total lands between 0 and 100. No LLM decides anything — fixed rules produce the number, which is why you can walk an LP or a partner through exactly how it was made.

Worked example — a seed deal

Take a seed company you rate: Team 4, Market 4, Product 3, Traction 3, Terms 4. Using the seed column:

That's 73 / 100. The bands are: 78+ Strong · 62–77 Promising · 45–61 Watch · below 45 Weak. So 73 is "Promising" — advance to a full memo. (This is the exact score in our fictional Halden Thermal sample memo; you can trace every line there.)

Red flags override the score

A high score is not a green light if something structural is broken. The scorecard carries a red-flag checklist, and flags don't just subtract points — they cap the recommendation:

• Fatal flags force the recommendation to Pass no matter the score. These are the items that quietly kill deals in confirmatory diligence: a messy or undocumented cap table, unresolved co-founder conflict, or IP not assigned to the company. An 85-scoring deal with unassigned IP still reads Pass — resolve first.
• Non-fatal flags knock the band down one step each (up to two), so an impressive-on-paper deal with, say, unverifiable metrics can't sail straight through. A "Strong" with one non-fatal flag becomes a "Watch."

Screen your own deal now

The full stage-weighted scorecard is live and free on the landing page — pick a stage, rate the five dimensions, toggle the red flags, and watch the score and recommendation update. No email needed to see your number.