The Money Slide Problem

Ask a founder what their deck is missing and they will usually say something about polish. Better visuals. Tighter copy. A stronger opening. Search for the most common pitch deck mistakes and you will find a hundred articles agreeing with them: fonts, wall-of-text slides, the 2x2 matrix with your logo conveniently in the top-right corner.

That is not what we found.

Between June and August 2026 we scored 34 pitch decks through the same rubric - ten dimensions, 0 to 100 each, applied identically to every deck. Pre-seed and seed, mostly European, weighted toward deeptech, AI and marketplaces. When we ranked the ten dimensions from worst to best, the bottom of the table was not about presentation at all.

Read the table from both ends and the shape of the problem appears. The dimension founders handle best is the one describing what they built. The three they handle worst all describe what it costs, what it returns, and what they are asking for.

Founders can explain their product. They cannot explain their money.

We checked whether that ordering was an accident of who happened to send us decks. One account in our sample was an accelerator that submitted a whole cohort, so we recomputed every finding with that account removed entirely. Financial projections stayed last. Solution stayed at the top. The middle of the table shuffled - but the two ends did not move.

The gap founders never see coming

The ranking tells you which slides score badly. A second cut tells you which content is simply absent.

For every deck, the analysis names the topics that should exist and don't. Sorted by how often each topic came up:

• Unit economics - named in 76% of decks. Cost to acquire a customer, what that customer is worth, when the two cross.
• Risk and regulation - 71%. What could go wrong, and what happens to the plan if it does.
• Traction and validation - 68%. Letters of intent, pilots, waitlists, anything that proves someone outside the building wants this.
• The ask - 56%. How much, and what it buys.

Unit economics tops that list by a clear margin, and it is the one founders are most surprised by. A deck can have a revenue forecast, a pricing slide and a market size, and still leave an investor unable to answer the only question that matters: does this business make money on each customer, and how do you know?

There is a second-order effect here. Most founders treat "financials" as one slide to be produced late, under time pressure, by whoever is least busy. So it gets a hockey stick and a headline number. The hockey stick is the part investors trust least, and the unit economics - the part that would justify it - is the part that never gets built.

The inconsistency is louder than the average

Averages hide the more interesting finding. Look at how tightly each dimension clusters.

Solution & value proposition has a standard deviation of 7. Almost every founder in our sample was decent at explaining their product; the scores bunch together in a narrow band.

Ask & use of funds has a standard deviation of 23 - more than three times as wide. A quarter of decks scored below 30 on it. Another quarter scored above 72.

That is not a bell curve. That is two populations. The ask is not a dimension founders are mediocre at - it is one they either take seriously or omit entirely. There is very little middle.

Financial projections behaves the same way: the middle of the pack sits at 48, but the bottom quarter is below 34. When a deck is weak on money, it is not slightly weak. It is empty.

You know your business better than your customer

Underneath the ten dimensions, the analysis runs a second pass on depth of understanding, across three lenses: how well the founder understands their own business, their customer, and the environment they are operating in.

Customer understanding came out lowest in 74% of decks - and, tested pair-by-pair within each deck rather than across the sample, it sits a consistent 4 to 6 points below the other two.

The margin is modest and we would not build a headline on its size. The consistency is what makes it worth reporting. Three quarters of founders demonstrated more command of their own business model than of the people they intend to sell it to. It shows up as the same absence every time: a customer described as a segment rather than a person, a problem asserted rather than quoted, and willingness to pay inferred from market size rather than from anyone actually paying.

That failure and the unit-economics gap are the same failure wearing different clothes. Both come from reasoning about customers in aggregate instead of one at a time.

What actually fixes this

The findings point at four concrete changes, in order of how much they move the score.

Build the unit-economics slide first, not last. One slide: what it costs to acquire a customer, what they are worth over their life, how long until they pay back. If you don't know these numbers yet, say so and show the assumptions you are testing. A stated assumption with a number beside it beats a confident blank.

Make the ask a sentence, not an implication. How much you are raising, what it buys, and what milestone it gets you to. Decks that omit it are not being modest - they are asking the reader to do work the reader will not do.

Replace the projection with the driver. A five-year revenue curve tells an investor nothing about whether you understand your business. The model underneath it does: how many customers, at what price, acquired through which channel, at what cost. Show the drivers and the curve becomes credible; show only the curve and it reads as decoration.

Quote one real customer. Not a persona. Not a segment. One named conversation with a person who has the problem, in their words. It fixes the customer-understanding gap and the validation gap in a single slide.

Notice what is not on this list. Nothing about fonts, colour palettes or slide count. Storytelling and design ranked fifth of ten in our sample - middling, and comfortably above everything financial. Design is the thing founders can see is imperfect, so it is the thing they fix. The money is the thing they cannot see is missing.

Method, and what we refuse to claim

Most articles on this topic are opinion. This one is data, which means it comes with limits, and we would rather state them than have you find them.

What the sample is. 34 companies, submitted by 12 separate accounts, analysed between 1 June and 20 August 2026. Pre-seed (20), seed (12), Series A (2). Sectors skew to deeptech, AI and marketplaces. Predominantly European. Where a company submitted more than once we used only its first submission, so no startup is counted twice.

What we removed. Every deck submitted from our own accounts, and our own company's deck. Two files that were not pitch decks - a one-page document and a three-slide executive summary. All analyses from before June 2026, for the reason below.

Why the window starts in June. Our scoring engine changed. We re-ran two byte-identical PDFs from the earlier period through the newer engine and one dimension moved by 27 points. Pooling the two periods would compare decks scored under different rules, so we did not. The earlier decks are excluded rather than adjusted.

How much the numbers wobble. Re-running the same PDF twice moves the overall score by a median of about 9 points. That is our noise floor. We do not report any difference smaller than it as a finding, which is why the middle of the ranking is described as shuffling rather than ranked.

Where the sample is thin. The 34 companies come from 12 accounts, and one accelerator contributed 12 of them. Companies inside one account resemble each other, so the effective sample is meaningfully smaller than 34. Every headline above was computed twice - with that account and without it - and only the findings that survived both cuts are reported here.

What we are not telling you. Several numbers that looked good did not survive that test, and they are not in this article. The share of decks with no identifiable ask fell from 29% to 9% depending on whether the accelerator cohort was included, so we report no figure for it. Nor do we report how often specific cognitive biases appear: the analysis is instructed to name several per deck from a short list, so a high percentage for any one of them is arithmetic, not evidence. We also make no claim about median deck length, because our sample mixes demo-day decks with investor decks and those are different objects.

What these scores are not. They are one rubric's reading of what a deck communicates. They are not investor decisions and not fundraising outcomes. A deck scoring 48 on financials is not a company that will fail to raise - it is a deck that leaves an obvious question unanswered.

We publish these limits because the alternative - a round number with no denominator - is what the rest of this genre already offers, and it is worth less than it looks.

The one-sentence version

Founders spend their preparation time on the half of the deck they are already good at. The three weakest dimensions in our sample were financial projections, the ask, and traction - and the strongest was the product. If you have an hour to spend on your deck this week, spend it on the money.