We Read 178 Founder Pitch Decks. Here Is What Was Missing

From August 18 to October 20, 2026, our Seed Partner Review read 178 pitch decks from 176 founders. It scores each deck against the eight questions a seed investor asks, then marks up every slide. Here is what the decks had in common, and what most of them left out.

Read this first. These are our own scores, produced by an AI model working from a fixed rubric, on decks that founders chose to upload to a funding-help site. They show how a deck reads against that rubric. They do not show whether an investor would fund the company, and we have not tracked whether any of these founders raised. The full method is at the end.

Founders explain the idea, then stop

Each deck gets a score from 0 to 100 on each of the eight questions. The idea itself was the one thing most decks got across: that question averaged 56. The other seven averaged between 26 and 33.

Part of the gap is blanks. When a deck never addresses a question, the review scores it low and marks it unanswered. 74 of the 178 decks never said why these particular founders are the ones to build this. 69 never addressed competition. 56 never explained how the company grows, 51 had nothing on what stops a rival copying it, and 50 showed no traction.

Across all eight questions, 139 decks left at least one unanswered and 59 left three or more. Only 39 answered all eight. Much of the gap is not weak writing on the slides that exist. It is questions that no slide answers.

The slides that make claims cost the most

Every slide gets one of four labels: keep, needs work, delete (it costs attention and returns nothing), or damaging (it makes an investor trust the company less). Across 2,426 slides, 33% were kept, 35% needed work, 16% were marked delete and 16% damaging. 156 of the 178 decks, 88%, had at least one damaging slide.

Cover and problem slides were the safest: 81% of covers and 56% of problem slides were kept as they were. The riskiest were the ones that ask an investor to believe a number or a person: financials, the team, the ask and traction. Our reading is that these are the slides where a claim needs proof, and the proof was missing or thin. We did not record why each label was given, so treat that as a reading and not a measurement.

If you are cutting slides, cut these first. Roadmap slides were marked delete 29% of the time, and thank-you or contact slides 27%.

Longer decks scored higher. That is not a reason to add slides

Score rose with length up to 24 pages, then flattened. Only six decks ran past 24 pages, so treat that last row with care. The most common length was 9 to 12 pages, and those decks averaged 27.

We read this as room, not reward. A deck has to answer eight questions, and a very short deck often has nowhere to put the proof. That does not make the usual advice to keep a deck tight wrong, because investors read quickly. It means a short deck that skips the market, the competition and the traction is not tight, it is incomplete. Add a slide only when it answers one of the eight questions, and cut the ones that do not.

Stage moves the score less than you would expect

Early-revenue decks scored highest, at 40. The later-stage decks, 28 of them, averaged 35, a little below. Having a bigger business did not, on its own, make the deck better.

The eight questions, as a checklist

Before you send a deck, check that some slide answers each of these. They are listed in the order decks most often left them blank, with the number of decks that did.

1. Why these founders (74 blank). What do you know about this problem that an outsider does not? Show specific experience, access or technical ability on the team slide, not just job titles.
2. Competition (69). Name the direct competitors, the indirect ones, and what customers do today, including nothing. Say why customers are not already using those.
3. Growth (56). How does the company grow, and does that mechanism get stronger as you get bigger? Word of mouth, sales, partnerships and paid ads all count, but say which.
4. Moat (51). "AI-powered platform" is not an advantage. Proprietary data, distribution, switching costs, regulation or hard-won know-how are.
5. Traction (50). Sign-ups are the weakest form. Show who pays, how often they come back, and whether the number is growing without you pushing it.
6. Market (30). Start with the first customer: how many of them exist and what each is worth. A top-down figure from a report is a weak substitute.
7. Business model (29). What does a customer pay, what does it cost you to serve them, and how long until they have paid back what it cost to win them? Early on, show that the economics could work.
8. The core idea (1). Who has the problem, why it hurts, and why now. Almost every deck already does this part.

For what to put on each slide, see our guide to creating a pitch deck, slide by slide.

How we got these numbers, and what they cannot tell you

• What was counted. Every finished read of a pitch deck between August 18 and October 20, 2026: 178 reads from 176 founders. Business plans and one-pagers are not included.
• How a deck is scored. An AI model (Google's Gemini) reads the file against a fixed rubric of eight questions, scores each from 0 to 100, gives an overall score and a band (not ready, needs work, close or strong), then labels every slide. A question the deck does not address is scored low and flagged unanswered, so the averages above include those decks.
• It is our rubric, not investors' decisions. We have not followed these founders to see who raised. A low score means the deck does not answer the questions we ask, not that the company is a bad bet.
• Who is in the sample. Founders who chose to upload a deck to a funding-help site. 107 of the 178 described themselves as idea or MVP stage, so the sample leans early. We did not record founder country for most of these decks, so we cannot say how they split by region.
• Slide types. Matched by keywords in each slide's title. About six in ten slides matched one of the types in the chart, and the rest are left out of it.
• Privacy. This post uses counts and averages only. No deck, company or founder is identified, and nothing is quoted from a deck.