This page sets out figures from a study of 116 startup files: 101 multi-page decks and 15 one-pagers, shared by exhibiting startups at Bits & Pretzels 2026. Most of the teams are at pre-seed or seed stage. 50 of the 116 decks pitch to investors, 40 sell to customers and the rest are company introductions or partner decks; figures that apply only to investor decks say so. On this page, "decks" means all 116 files, including the 15 one-pagers, unless a figure names a narrower group. Each figure is given with the base it is counted from. By Derrick Cramer, Founder, Gossamer.
Who the decks are for
- 50 of the 116 decks (43%) pitch to investors, including 3 that also pitch customers or partners.
- 40 of the 116 decks (34%) sell to customers, 22 (19%) are a general company introduction and 4 (3%) address partners.
- 36 of the 50 investor decks state how much they are raising. The 34 with a figure that could be extracted ask for a median of EUR 1.8M, ranging from EUR 120K to EUR 15M.
- Where an investor deck names its round, it is pre-seed (12 decks) or seed (12) far more often than Series A (4) or growth (4).
- 39 of the 50 investor decks show how the money will be spent.
- 69 of the 94 decks that carry a date were dated August or September 2026, and 12 decks mention Bits & Pretzels by name.
Investor decks are the largest group, with customer decks close behind
Base: 116 decks. Source: Gossamer, November 2026.
What the decks contain
- The median multi-page deck has 13 slides.
- 61 of the 100 decks of five or more pages open on the problem, 11 on an executive summary, 8 on a vision statement and 6 on the team.
- 51 of the 100 decks of five or more pages close on contact details or a call to action, 13 on "thank you" and 10 on the ask.
- Investor decks of five or more pages cover a median of 7 of the 9 core slide types, and largely keep the classic order: the median rank correlation with an adapted Sequoia outline is 0.79, where 1.0 is the identical order.
- 74 of the 116 decks have a team section, 34 have a customer-logo section and 15 include a case study.
- PDF metadata confirms the creation tool for 69 of the 116 decks: PowerPoint for 23, Canva for 17 and Google Slides for 12. PowerPoint accounts for between 20% (confirmed) and 30% of all decks; the upper end allows for up to 12 decks with no tool metadata that use PowerPoint's widescreen page size.
How startups describe themselves
- AI is the core of the product in 50 of the 116 decks (43%), a feature in 15, mentioned in passing in 9 and absent in 42.
- Of 86 category phrases, 20 (23%) contain "AI" and 18 (21%) contain "platform", the two most frequent category words tracked.
- 31 of the 116 decks never state a category. 58 use a plain category label, 14 use "X for Y" and 13 a borrowed metaphor such as "the OS for".
- 97 of the 116 decks (84%) describe their customer as a segment, and 9 (8%) name the role that makes the purchase. Another 9 stay generic, and 1 names no customer.
- 67 of the 116 decks (58%) target customers across industries.
- 87 of the 116 decks (75%) frame the problem as an operational pain the customer feels.
- 53 of the 116 decks explain why the timing is right. They cite market or cost shifts (34), technology (20) or regulation (19).
- 34 of the 116 decks (29%) name a regulation or standard, led by GDPR (24) and the EU AI Act (13).
AI is the core of the product in 50 decks and absent from 42
Base: 116 decks. Source: Gossamer, November 2026.
Most decks describe their customer as a segment, and few name the buyer
Base: 116 decks. Source: Gossamer, November 2026.
Competition and defensibility
- 46 of the 116 decks (40%) do not address competition in any form.
- Decks that address competition name categories of competitor (37), name rivals (36), contrast with the status quo (22) or use feature tables (21); many use more than one of these. 9 use a 2x2 positioning grid.
- 69 of the 116 decks (59%) claim no defensible advantage. The rest cite technology or IP (31), data (22) or switching costs (11).
Two in five decks skip competition, and three in five claim no defensible advantage
Base: 116 decks; a deck can use several approaches. Source: Gossamer, November 2026.
Business model
- 91 of the 116 decks are B2B.
- Subscription is the most common revenue model, in 51 of the 116 decks.
- 41 of the 116 decks (35%) never make the revenue model clear, and 42 show prices.
Proof
- 21 of the 116 decks (18%) cite no proof of any kind.
- 16 of the 116 decks (14%) cite revenue or ARR.
- The most common forms of proof are usage metrics (42 decks), customer logos (38), pilots or letters of intent (31) and customer counts (27).
- 19 of the 47 investor decks of five or more pages (40%) show no traction metric. Of the same 47, 5 leave out market size, 4 the ask and 2 the problem, and none leaves out the solution or the team.
- Of the 32 customer decks of five or more pages, 8 offer no proof of any kind, 7 show no product and 6 never state the problem.
- 42 of the 116 decks (36%) make at least one "leading", "first" or "only" claim with no number or name attached.
- 61 of the 74 team sections give credentials, most often years of domain experience (45). Prior exits are rare (5).
Usage metrics and customer logos are the proof decks cite most
Base: 116 decks; a deck can cite several kinds of proof. Source: Gossamer, November 2026.
Traction is what investor decks leave out most often
Base: 47 investor decks of five or more pages; each bar counts the decks missing that part. Source: Gossamer, November 2026.
Writing
- The median deck carries 77 words per slide (100 decks of five or more pages).
- Across 85 English multi-page decks with 300+ words, the median Flesch reading-ease score is 28, and 47 of the 85 score below 30; plain-English guidance aims for 60 or more. Slide text is fragmentary, so these scores are approximate and best read as a sign of dense text.
- 50 of the 85 decks use at least one word from a fixed buzzword list, such as "seamless", "unlock" or "empower".
- Across the same 85 decks, about two thirds of these pronouns are "we" or "our" (median share 0.67), so startups write about themselves about twice as often as they use "you" or "your".
Design and legibility
- 43 of the 116 decks (37%) use a dark background, 67 a light one and 6 a mix.
- 66 of the 116 decks (57%) lead with a blue, cyan or violet accent colour; warm accents lead in 16.
- In 54 of 102 measurable decks (53%), at least 5% of the text is set below 10 pt on a slide normalised to 7.5 inches high (8 pt for portrait A4), counting footnotes and captions.
- 16 of 102 measurable decks (16%) fall below the 4.5:1 accessibility contrast ratio.
- 67 of the 116 decks (58%) include at least one chart, with 188 charts across 1,372 slides.
Just over half of measurable decks set some text below 10 pt
Base: 102 measurable decks. Source: Gossamer, November 2026.
Differences between groups
- 3 of the 50 investor decks (6%) show no proof at all, against 12 of the 40 customer decks (30%).
- 10 of the 50 investor decks (20%) show no competition, against 18 of the 40 customer decks (45%).
- 52% of decks from teams of 11 or more people (n = 31) have a customer-logo section, against 21% from teams of 1 to 10 (n = 85).
- 12 of the 15 one-pagers show no competition.
Customer decks skip proof and competition far more often than investor decks
Base: 40 customer decks and 50 investor decks. Source: Gossamer, November 2026.
These differences are descriptive and were not tested for significance. The split by audience rests on coding with moderate agreement (see below), so treat it as indicative.
Method and limits
The files are the ones exhibiting startups attached to their booth listings on the event's app. 123 of the 230 Startup Expo exhibitors at Bits & Pretzels 2026 uploaded something. After removing corporates, service providers, a duplicate and a damaged file, 116 startup files remain. 106 are in English, 9 in German and 1 mixed.
Scripts measured the creation tool, colours, text size, contrast and wording. An AI model (Claude Sonnet) coded every deck against a codebook that was tested on a 10-deck pilot and frozen before the full run; 9 of the 10 pilot decks are among the 116 and were re-coded under the final rules. Each judgement cites its slide. A second model (Claude Opus) re-coded 10 random decks as a check. Agreement was 89% on slide labels (Krippendorff's α 0.88), 96% on components present (α 0.93), between α 0.70 and 1.00 for most deck-level judgements, and 80% on audience (α 0.67, below the usual 0.70 bar). With only 10 decks checked, every agreement figure carries wide uncertainty. No person validated the coding. Agreement between two AI models shows the coding is consistent, and it does not prove the coding is correct.
The sample describes exhibitors at one event: 85 of the 116 (73%) come from teams of 1 to 10 people, mostly at pre-seed or seed. Some decks are sales handouts, which is why figures are split by audience. Several sub-samples are small (47 investor decks and 32 customer decks of five or more pages, 15 one-pagers), and the 9 German decks are counted without being compared as a group. Whether a deck was built with an AI assistant cannot be told from the file, so no figure here measures that.
Cite as
Cramer, D. (2026). Startup pitch deck statistics 2026: 116 decks from Bits & Pretzels. Gossamer. https://www.gossamergrowth.com/pitch-deck-statistics
The figures on this page may be reused with credit to Gossamer and a link to this page.
