The One-Page Memo That Outperforms the Deck

A one-page memo beats a deck because the decision that matters gets made in a room you are not in, by people reading without you, inside an attention budget of under three minutes. The memo is the only artifact that survives that trip with its argument intact.

A deck is a visual aid for a narrator. Strip the narrator and most of the payload goes with them. The room where your round gets decided has no narrator.

Write the page. Send the page. Build the deck when somebody asks for it.

Why these three models

This piece runs the Wire Model: score the features of the decision, route to a small ensemble of formal models, then force dated actions out of it. The scores that mattered:

• Cognitive and affective distortion (0.85). Attention is rationed in seconds, and delivery moves belief in directions unrelated to the business.
• Strategic actors with asymmetric information (0.8). You know where your argument is thin. The format either exposes that or hides it.
• Transmission topology (0.75). The artifact must replicate through a partnership, an angel network or a committee, with no author attached.
• Regime-break risk (0.7). Machines now read what you send, and machines now write it. Both ends of the signal are moving.
• Deep uncertainty (0.5). Fund timing, mandate drift and committee politics are invisible from where you sit.

That routes to three models: contagion (whether the document replicates), signaling (what the format costs you to produce), and behavioral (what a reader’s attention does to it on arrival). They span three outcome types, complex, equilibrium and noise, so their errors point in different directions and partly cancel.

Mechanism design scored well and got cut. Contagion already hands you its lever, which is writing the investor’s own memo for them, and a fourth card costs about 350 words the argument needs more.

The framework: your raise is a document that has to travel without you

1. Transmission: the artifact that decides is a memo, and somebody else writes it

Start with the document that actually decides. When Kaplan and Stromberg studied how venture capitalists choose, they did not watch pitches. They obtained the written investment memoranda partners prepare for their own committees, covering 90 investments in 67 portfolio companies across 11 firms, and reconstructed how each firm weighed the business, the risks, the management and the deal terms.¹ That is a methodology note. It is also the answer. Writing that memo is how the money gets approved.

The document that decides your round is a memo, and somebody else writes it. Your only real move is to write their first draft.

DocSend’s telemetry puts a deck review at under three minutes. At pre-seed in 2020, decks that went on to raise held attention for over four minutes, while decks that failed got one minute thirty.² Causality runs both ways, so read the floor rather than the gap: a deck that cannot hold ninety seconds dies before anyone forms a view.

The same dataset puts a typical deck at roughly twenty pages carrying about fifty words each.² That is a thousand words of fragments with the connective tissue stripped out, and the connective tissue is where your reasoning lives. You carry it in your head and supply it live, which works exactly once, to one person. Then the partner forwards the file on Friday night to two colleagues who have never heard of you. A memo replicates at close to full fidelity. A deck loses the argument and keeps the logos.

Outside the largest hubs, that forward is the entire process. Organised angel networks now run across 37 African countries, and about 80% of their deals still land in Nigeria, Egypt, Kenya and South Africa, so the money reaching a founder in Lusaka or Dakar is mostly local and diaspora capital, pooled by people in other time zones who read you on a phone, in a thread, on a Sunday.³ African companies disclosed US$3.8 billion across 2025, with fewer than 5% of deals crossing US$50 million while taking about half the capital.⁴ Your round sits in the long tail, where nobody is flying in for a presentation.

Contagion, the transmission lens

Assumes: the document gets forwarded and must carry its own meaning when it does.

Fits because: transmission topology scored 0.75; the deciding artifact is a memo written by someone absent from your meeting.

Breaks when: one person writes the cheque alone. Solo angels and family offices decide in the room, and there the live pitch keeps its power.

Counteracts: optimising for the meeting you can see instead of the one you cannot.

May reinforce: document fetishism, and the belief that a good page can substitute for a business.

2. Signaling: prose is expensive in the one way that matters

Spence’s result is the cleanest lens here. A signal separates types only when it costs something, and only when it costs the weak sender more than the strong one.⁵ Slide-making costs everyone about the same. Writing does not.

Amazon runs the experiment at scale and publishes the result. It replaced presentations with narratively structured six-page memos, and Jeff Bezos was explicit about the cost: great memos “are written and re-written, shared with colleagues who are asked to improve the work, set aside for a couple of days, and then edited again with a fresh mind. They simply can’t be done in a day or two.” His guidance was that one should take a week or more.⁶ That cost lands unevenly on purpose. A founder who understands their own unit economics compresses them in an afternoon. A founder running on vibes sits in front of the page for three days and it still will not close.

People believe they understand mechanisms far better than they do, and the illusion collapses the moment they are asked to write out a step-by-step causal explanation. Rozenblit and Keil found the effect far stronger for explanatory knowledge than for facts or procedures.⁷ A bullet is a fact. A paragraph is an explanation. Only one of them exposes the gap.

A slide can hold a missing causal chain. A sentence cannot.

The most expensive recorded instance sits in the Columbia Accident Investigation Board report. Engineers presented a debris-strike analysis on slides. The critical number, one debris estimate several hundred times larger than anything used to calibrate the model, sat at the bottom of the page under a headline saying the opposite. The Board’s conclusion, on page 191: “it is easy to understand how a senior manager might read this PowerPoint slide and not realize that it addresses a life-threatening situation.” It named “the endemic use of PowerPoint briefing slides instead of technical papers” as an illustration of NASA’s problematic methods of technical communication.^{8, 9}

Your burn rate threatens nobody’s life. The mechanism is identical. Bullet hierarchy hides the number that decides.

Signaling, the cost lens

Assumes: the cost of producing the artifact tracks the quality of the thinking behind it.

Fits because: asymmetric information scored 0.8, and writing prices clarity in a way slide-making does not.

Breaks when: generation is free. A language model writes a causal-sounding page for a business that has none, which is the regime break below.

Counteracts: the belief that presentation polish is evidence of anything.

May reinforce: penalising founders writing in a second or third language, and mistaking fluency for operating skill.

3. Behavioral: the attention budget and the delivery premium

Do the arithmetic before arguing about taste. Silent reading of English non-fiction runs at about 238 words per minute, averaged across 190 studies and 18,573 participants. Reading aloud runs at 183.¹⁰

A twenty-page deck at fifty words a page is a thousand words: four minutes and twelve seconds of reading against an observed budget of under three. It does not fit, so the reader skims, and skimming a deck means looking at pictures.

A five-hundred-word memo reads in two minutes and six seconds. It fits with room to spare and every word lands. That is the whole case, and you can redo the arithmetic with your own deck in a minute.

The live pitch is slower still: 183 words a minute, and it costs a calendar slot from someone with no reason to give you one. It also carries error. Hu and Ma processed 1,139 startup pitch videos with machine learning, scoring visual, vocal and verbal persuasion. Positive, passionate, warm delivery raised the probability of getting funded. Conditional on funding, those same high-positivity startups underperformed, and a follow-on experiment showed why: delivery works by leading investors to form inaccurate beliefs.¹¹

Charm raises your odds of getting funded and lowers the odds you deserved it.

The hybrid you probably run is the weakest of the three. Narrating slides that repeat their own text degrades retention and transfer relative to narration alone, across four experiments by Mayer, Heiser and Lonn.¹² Reading your bullets aloud costs the room comprehension.

Behavioral, the attention lens

Assumes: attention is a hard budget and readers deviate from careful evaluation in stable directions.

Fits because: distortion scored 0.85; the budget is under three minutes and delivery demonstrably moves belief.

Breaks when: the screen is scored against a written rubric by a named reader. Structure removes most of the effect.

Counteracts: the assumption that a good business gets read carefully because it deserves to be.

May reinforce: writing for skimmers until nothing substantial is left to skim.

GEER: the levers, cheapest first

Four channels carry everything you can change. Legibility: can a stranger restate it. Verifiability: can they check it without you. Transmissibility: does it survive a forward. Attention cost: does it fit the budget.

1. Write the page. Under 600 words, seven parts: what you sell and to whom, in the customer’s words; who pays and how much; the one number that moved and over what period; why now; what the money buys and by when; the risk you would flag from their side; the ask, with amount, instrument and commitments to date. Costs an afternoon.
2. Move the ask into the first paragraph. By the end of paragraph one a reader should know the amount, the instrument and how much is already circled. Hits attention cost.
3. Replace every adjective with a number and a source. “Strong retention” is noise. “Month-six agent retention 61%, from the operator’s settlement export, link below” is checkable in ten minutes. Hits verifiability.
4. Kill the attachment. Put the memo in the body of the email. Attachments do not survive a forward into a thread, a network digest or a WhatsApp group. Hits transmissibility.
5. Run the restatement test. Hand the page to someone outside your sector and have them tell you what you sell, to whom, and what you want. If they miss, the page is wrong. Hits legibility. Costs a day.
6. Build the verification appendix. One link per number: the signed LPO, the mobile-money settlement statement pulled from the operator, the regulator’s letter, the customer who agreed in writing to take a call. Hits verifiability. Costs weeks.

No-lever flag: if the page refuses to close because a step in the causal chain is missing, you are in a proof-required state. No artifact fixes that. Write that down and re-plan the quarter around producing the missing proof.

RADAR: the portfolio, dated

DO NOW, by T+3 days. Reversible and dominant across every scenario.

1. Write the memo. Under 600 words, ask in the first paragraph, three numbers a stranger can verify.
2. Run the restatement test on one person outside your sector. Rewrite until they pass.
3. Send it to six named investors in the body of an email. Six, not sixty.

HEDGE, by T+14. Cheap insurance against the page being read badly or not at all.

1. Build the verification appendix as one linked page. Nobody has to open it, and its existence changes how the memo reads.
2. Paste the memo into a language model and ask what the business does, who pays, and what the ask is. If it comes back wrong, a screening tool will too.
3. Keep an unformatted plain-text version. That is the one that survives every forward.

DEFER AND TRIGGER. Expensive or irreversible, so wait, and pre-commit the trigger now.

1. Defer: the full deck, the data room, the pitch video, the paid design.
2. Trigger to build: an investor asks, or two of your first six memos convert to a call by T+14. Then build the deck from the memo rather than the other way round.
3. Counter-trigger: zero of six by T+28. The gap is in the proof. Stop sending and go build the certificate.

If you are the one being asked. DO NOW: require a one-page memo before any first meeting and publish that rule. HEDGE: write your own memo before the call, so the call tests your thesis instead of forming it. DEFER: changing your screening model until two quarters are measured on memo to first meeting, first meeting to term sheet, and 24-month survival.

CHAIN: what usually happens next

Pick the reference class by structure and ignore the surface. The structure here: a consequential decision taken by a group, from a document, with the author absent. Venture investment committees run it. So do NASA flight-readiness reviews and Amazon’s operating meetings. The base rate across all three points one way. When the author is gone and the stakes are real, the narrative document beats the slide.^{1, 6, 8} Present-state modifiers push the same way: attention budgets keep shrinking, screening keeps moving cross-border and asynchronous, and angel networks in 37 countries now pool decisions in group threads.^{3, 2}

Subtract the counterfactual before over-crediting the format. Part of the memo’s advantage is selection: founders who can compress a business to one page usually have a clearer business and would have raised anyway. The format does less work than the correlation suggests. It remains your cheapest intervention this week.

Matrix-break flag. Two things moved at once. Machines now read your submission: an LLM screening agent worked through a database of 61,814 early-stage companies, clustering at quality comparable to human analysts while running 537 times faster.¹³ Text feeds that pipeline losslessly. A slide deck does not. Machines also write it now, stripping prose of its separating cost, since anyone can produce a causal-sounding page in a minute. Short run, text wins on both counts. Medium run, writing quality separates nobody, and the separator moves to specifics a third party can falsify. Build for the medium run: every claim in the memo checkable by a stranger without your help.

What this ensemble cannot see

Three gaps, each big enough to swallow a quarter.

Fund timing and mandate. Dry powder, cycle position, a mandate that quietly narrowed last quarter. None of it is visible from where you sit, and it often decides everything. Founders read that silence as a judgment on the company.

Whether the page was read at all. Plain text in an email body gives you no telemetry. You buy survivability and pay in analytics, so you will never know which of the six opened it.

The reputation the document arrives with. The same page from a second-time founder reads differently, and nothing here separates the memo’s quality from the name on it.

Compression and verifiability sit inside your control. Timing and reputation sit outside it. Spend the week on the first pair and price the second as weather. By T+3: one page, under 600 words, the ask in the opening paragraph, three numbers a stranger can check without you in the room. Send it to six. Leave the deck unbuilt until one of them asks.

Sources and notes

1. Kaplan, S. N., and Stromberg, P. “Characteristics, Contracts, and Actions: Evidence from Venture Capitalist Analyses.” Journal of Finance 59(5), 2004, 2177-2210. The study analyses the written investment memoranda VC partners prepared for their investment committees, covering 90 investments in 67 portfolio companies by 11 venture partnerships between 1987 and 1999. Working paper: NBER w8764 (full text).
2. DocSend pitch deck research, published by Dropbox. Reports that investors spend under three minutes reviewing a deck; that in 2020 investors spent over four minutes on pre-seed decks that raised versus one minute thirty on decks that did not; and that most decks run roughly twenty pages at about fifty words per slide. Research page. Aggregated platform telemetry rather than a controlled study; the raised-versus-not comparison is correlational.
3. African Business Angel Network (ABAN), 2025 Angel Investment Survey Report, produced with Briter. 60+ responses from active angels and angel networks across Africa. Reports organised angel networks operating across 37 African countries, roughly 80% of angel deals concentrated in Nigeria, Egypt, Kenya and South Africa, and the mobilisation of local and diaspora capital. Full report PDF.
4. Briter, Africa Investment Report 2025. US$3.8 billion disclosed across 2025, announced deals up 8% year on year, fewer than 5% of deals above US$50 million accounting for about half of disclosed funding. Report page.
5. Spence, M. “Job Market Signaling.” Quarterly Journal of Economics 87(3), 1973, 355-374. Full text.
6. Amazon, 2017 Letter to Shareholders (Jeff Bezos). Source of “We don’t do PowerPoint (or any other slide-oriented) presentations at Amazon. Instead, we write narratively structured six-page memos,” and of the passage on memos taking a week or more. Full letter.
7. Rozenblit, L., and Keil, F. “The Misunderstood Limits of Folk Science: An Illusion of Explanatory Depth.” Cognitive Science 26(5), 2002, 521-562. Full text, Yale Cognition and Development Lab.
8. Columbia Accident Investigation Board, Report Volume I, August 2003, page 191, section “Engineering by Viewgraphs.” Source of “it is easy to understand how a senior manager might read this PowerPoint slide and not realize that it addresses a life-threatening situation” and of the Board’s finding on “the endemic use of PowerPoint briefing slides instead of technical papers.” Full report PDF.
9. Tufte, E. R. Analysis of the Boeing debris-assessment slide, reproduced and cited by the Board on the same page. Tufte’s reading of the buried debris-size estimate and the misleading headline is the basis for the Board’s commentary. Author’s notebook entry.
10. Brysbaert, M. “How Many Words Do We Read Per Minute? A Review and Meta-Analysis of Reading Rate.” Journal of Memory and Language 109, 2019, 104047. Silent non-fiction reading 238 wpm across 190 studies and 18,573 participants; oral reading 183 wpm across 77 studies and 5,965 participants. Full text PDF.
11. Hu, A., and Ma, S. “Persuading Investors: A Video-Based Study.” NBER Working Paper 29048, 2021; published in the Journal of Finance, 2025. 1,139 startup pitch videos processed with machine learning. Positive delivery raises funding probability; conditional on funding, high-positivity startups underperform; an experiment shows delivery works by leading investors to form inaccurate beliefs. Abstract. Author PDF.
12. Mayer, R. E., Heiser, J., and Lonn, S. “Cognitive Constraints on Multimedia Learning: When Presenting More Material Results in Less Understanding.” Journal of Educational Psychology 93(1), 2001, 187-198. Four experiments; adding on-screen text duplicating a narration lowered retention and transfer. Record and abstract.
13. Vismara, S., Latifi, G., Meinzinger, L., and Pass, A. “Generative AI-Powered Venture Screening: Can Large Language Models Help Venture Capitalists?” International Review of Financial Analysis 109, 2026. Database of 61,814 early-stage companies; the LLM agent operated 537 times faster than a human VC at comparable clustering quality. Record and abstract.