We built a managed agent called Alex that grades startup pitches on how fundable they look. The obvious risk with a tool like that is hindsight. Show it Airbnb's deck and of course it says yes, because everyone in the room already knows Airbnb became Airbnb. A grader that only flatters winners after the fact is worth nothing to a founder raising next week.
So I ran a cleaner test. I had Alex score five real decks exactly as they stood on the day they were sent, ignoring everything we now know happened since. Four of the names you will recognize. One you almost certainly won't, and that gap turns out to be the most useful part.
The two everyone knows
Airbnb's 2008 seed deck scored a 3.80 out of 5. A lean yes, not a slam dunk. The model liked the clear marketplace, the obvious revenue model, and the enormous market. It marked down the one thing that was genuinely weak: the deck argued demand by analogy, borrowing numbers from Craigslist and Couchsurfing instead of showing its own. That is exactly why a number of real investors passed on Airbnb early. The company went on to raise about $600K from Sequoia, and it is now a public company worth roughly $80 billion. A grader that returned "obvious fund" on that deck would have been lying, because the proof was not on the page yet.
Uber's 2008 deck scored a 3.61, for almost the opposite reason. There was no traction at all, because the company had not launched. What carried it was a near perfect problem slide (taxis in 2008 really were broken, and the deck said so specifically) and a timing argument that has since become a textbook example: smartphones with GPS were the new thing that made the whole idea possible. At the earliest stage, a five on the problem and a sharp "why now" can carry a deck with nothing running. It did here.
Notice that neither famous winner scored above a 3.8. That is the rubric refusing to peek at the ending.
The one that proves the point
Buffer's 2011 seed deck is my favorite of the five, because it does something almost no early deck does. It leads with proof. Eight hundred paying users, a $150K revenue run rate, 55,000 users growing 40% a month, 97% margins, and a freemium model with the conversion, churn, and lifetime value all written down. That earned a 5 on traction, and an overall 3.54.
What kept it a lean yes rather than a clear fund was scale. Social scheduling is useful and sticky, and it might also be a good business rather than a huge one. Our market dimension flagged exactly that. Buffer went on to raise a $3.5M Series A, turn profitable, buy out its investors, and stay independent. By any founder's measure it is a success. It is also not a unicorn, and it never tried to be. The scale flag was not a criticism. It was a forecast, and it was right.
The two you don't see coming
Mattermark raised a $6.5M Series A in 2014 on a strong revenue and growth story. Alex scored the deck a 3.31, a hair below the lean yes line, a genuine coin flip. The reasons were specific: real but modest traction for a raise that size, propped up by a steep projection, in a crowded market whose ceiling was unclear. Investors were more generous than our score and wrote the check. The company could not sustain the projected trajectory, struggled to raise again, and announced its sale to FullContact on December 21, 2017.
The fifth deck belongs to Usetrace, a browser-testing tool you have probably never heard of. That is the point. It scored a 3.02, borderline. The visible traction was real, 65 paying companies and $14K in monthly revenue, but the dimensions that decide whether a low-price tool becomes a venture-scale business, the market size and the moat, were thin. It raised a €120K seed on the strength of the traction. Then it shut down.
Four of these companies are household names. One is not. The thing that separates them is not how much money they raised, because all five raised. It is closer to what our market and scale dimensions keep poking at, and it is mostly invisible on the day the check clears.
Current view, subject to change
I want to be careful about what this proves. The weights behind these scores are hypotheses, not laws, and a sample of five decks read in hindsight is a story, not a study. We are calibrating Alex properly against a working venture fund's real decisions, plan by plan, and logging every time the model and the investor disagree. If the disagreements cluster, the weights move. What would change my mind fastest is a run of decks where the rubric's scale flag fires and the companies thrive anyway. So far that pattern has not shown up. I am watching for it.
Final thoughts
No one can look at a seed deck and tell you which company becomes a giant. Anyone who claims otherwise is selling something. What you can do, and what I think is actually useful, is read a document honestly. Here is what it proves, here is what it only asserts, and here is the one weakness most likely to matter. Do that before the market's verdict and the company's fate are known, and you are doing the same work a good investor does in the first ten minutes.
That is the whole job. Not predicting the ending. Reading the page in front of you, cold, and saying what is really there.
You can read all five evaluations, scorecards and all, at coworkers.global/examples.
Best regards,
Charles Stack
Founder, Coworkers.Global
Research by Alex, a managed startup-advisor agent at Coworkers.Global.
This article discusses public pitch decks and general fundability concepts. It is not a valuation opinion or financial advice.