This is the rubric we use to evaluate every business plan and pitch deck that comes through our fundability review. We are publishing it because founders should know exactly what they are being graded on, and because most advice on “how to evaluate a business plan” is either a listicle or a vibe. This is neither. It is the working methodology behind every review we deliver.
How scoring works
Every document is scored on eight dimensions, each rated 1 to 5 against written anchors, based on evidence in the plan itself. Two rules do most of the work:
Evidence or it didn’t happen. Every score cites the page or section it came from, or names what is missing. Absence of evidence is a low score, not a neutral one. A plan that never mentions competition has not avoided the question; it has answered it badly.
The document is what we score. We evaluate the plan as presented, not the business it might become. Those are both useful judgments, but letting one launder the other is how bad decisions get made.
The weighted scores roll up to one of four verdicts: Fund, Lean yes, Borderline, or Pass. A Lean yes means fundable with specific, addressable fixes. A Borderline means real gaps that could close. We publish the labels, not the numeric cutoffs, for the same reason we do not publish a fixed weight table (more on that below).
The eight dimensions, in order of weight
The order matters more than any single definition. This is the ranking, heaviest first:
1. Team and founder-market fit. The founders have lived the problem or shipped in this space, can build and sell what is proposed, and name their gaps with a hiring plan. Investors fund people who are unusually suited to a specific problem, and plans rarely fake this well.
2. Traction and evidence of demand. Real proof that someone wants this: revenue, retained users, signed pilots, a waitlist that converts. Not “interest.” Interest sits at the bottom of the proof hierarchy: interest, then revealed commitment, then retention, then revenue. Most plans claim the first and hope you read it as the last.
3. Problem and urgency. An acute, frequent, expensive pain for a specific buyer. A painkiller, not a vitamin. You should be able to say who hurts and how much it costs them.
4. Market and timing. Credible bottom-up reasoning about who you can actually reach, and a sharp answer to “why now.” We judge scale directionally, venture-scale or capped or unclear, rather than pretending a spreadsheet TAM is knowledge.
5. Solution and differentiation. Solves the stated problem materially better than the status quo, and the difference is defensible and obvious to the buyer, not just to the founder.
6. Business model and unit economics. Clear how money is made, pricing tied to value, and unit economics that work directionally or have a credible path.
7. Competition and defensibility. An honest map of the alternatives, including “do nothing,” and a real reason you win and keep winning.
8. The ask and use of funds. The amount tied to the specific milestones it buys and the runway it delivers. A round number with no plan is a finding, not a formality.
Two things to notice about the ranking. The top two dimensions together outweigh the bottom four combined, and usually by a wide margin. And the bottom of the list is not “unimportant”: a fatal flaw in any dimension can trip a gate, which brings us to the part of the rubric that overrides everything else.
Stage moves the weights
What we expect from a plan depends on where the company is, and we proxy stage by the size of the ask. Directionally:
- Very early (pre-seed): team and problem carry the bet. There is little traction to judge, so we do not pretend otherwise. Traction weight drops; team and problem rise.
- Seed: the base case. The ranking above.
- Series A and beyond: traction and unit economics become the expected proof and dominate. A Series A plan arguing from team alone is answering last year’s question.
Same dimensions, same anchors, different weights. A pre-seed plan is not graded on a curve; it is graded on the right test.
The red-flag gates
A great deck cannot outscore a fatal flaw. Five gates cap the verdict no matter what the weighted score says. One triggered gate caps the result at Borderline; two or more cap it at Pass. Every triggered gate is named in the output, with the evidence.
- No demand signal. Zero evidence anyone wants this. Only conviction.
- Numbers don’t reconcile. Projections, market size, or unit economics that contradict each other or appear fabricated.
- Uninvestable structure. A lifestyle-ceiling business presented as venture, with no plausible path to a return that justifies the ask.
- Team can’t execute. The team cannot build or sell what is proposed and has no plan to close the gap.
- Material dishonesty or omission. Misleading claims, or a critical section conspicuously absent.
In practice, the gates catch more real-world failures than the weighted score does. WeWork’s 2014 deck scores as Borderline on our rubric not because the vision was small but because the numbers did not reconcile, and that gate fired.
What every review outputs
Eight dimension scores with cited evidence, an overall verdict, the three highest-leverage fixes, the single biggest risk, a predicted invest-or-pass call, and a directional estimated valuation range. The valuation is always a dated range with a stated confidence level, never a single number, framed as what the market would likely support rather than what you are worth. Thin evidence shows up as a wide band, not a hidden assumption. None of it is a valuation opinion or financial advice.
Why the ranking is fixed and the percentages are not
We publish the exact weights we used on every deck we score in public. The Airbnb read shows them, and so does the WeWork read, and they are not the same numbers: 20% on traction for a 2008 seed deck, 28% for a 2014 Series D. That is the stage adjustment above, doing its work in the open.
What we do not publish is a single canonical table, because there isn't one. The weights are hypotheses under live calibration. For every plan we score, we log our predicted invest-or-pass against what the sourcing investor actually decided, and we tune against real outcomes rather than intuition. Freezing a table would misrepresent numbers that are deliberately still moving, and it would invite founders to optimize the document against the test instead of fixing the business. The ranking is stable. The decimals are a snapshot, and we date them.
We also score our own calibration in public. We have run the rubric blind against historical decks where the outcome is known and published the results, including where the rubric would have been wrong: Airbnb’s 2008 seed deck earned a Lean yes with the exact weakness real investors passed on, and WeWork’s 2014 documents tripped a gate we predicted would kill the round; the market funded it anyway, and we logged the miss in public. Demonstrated judgment, not claimed outcomes.
Get your plan read against this rubric
Alex, the managed agent we built at Coworkers.Global, runs this exact rubric on your plan: eight scores with cited evidence, the verdict, the three highest-leverage fixes, and the single weakest part to fix first. Every application gets a free screen: your fundability score and the weakest part of your plan. The full signed read is $295, and you decide after you have seen your screen. Drop your plan at coworkers.global/ai-business-plan-review. We are early and pre-revenue, so we lead with the quality of the read rather than a customer list we do not yet have. If you think the read is wrong, tell us.
Regards,
Charles Stack