Why strong technology loses the pitch

Iryna Manukovska · Fractional CMO, AI strategy and deep tech go-to-market · Storytelling workshop lead, AI House Deep Tech Accelerator · About

Published 7 September 2026 · Updated 7 September 2026 · 11 min read

Deep tech founders get offended when I ask them to compress forty slides into one page. "It can't be simplified, all details are critical." Investor narrative is a compression problem: you have to make the commercial consequence intelligible to a generalist while creating a credible path into technical diligence for a specialist. That needs two layers, not a shorter deck. This article also names five widely repeated pitch claims that have no traceable source, including two I used myself. Talk it through on a 30-minute call

Two-layer deep tech investor narrative: a decision layer covering customer, pain, why now, evidence and ask, and a diligence layer covering architecture, validation data, IP, regulatory pathway and manufacturing

What is the actual communication problem in a deep tech pitch?

You are explaining a system that needs months of diligence, in a meeting designed for rapid screening.

Paul Graham's framing has lasted because it is accurate: people who are bad at explaining, talking to people who are bad at understanding.

Judging three competitions across the UK and CEE last year and watching more than ten others, I saw one failure repeatedly. Founders put every idea and the entire feature backlog into a 180-second pitch. A fog machine, not a flashlight. Meanwhile the buyer-side data shows how little room there is for that: 79% of enterprise buyers complete their research before the first sales call, and 73% of decision-makers say they trust thought leadership over marketing materials. The room has already formed a view, and your job is to make the evidence legible rather than abundant.

There is a second-order problem specific to scientists. Scientists are trained in logical-scientific communication, which aims to provide general truths judged on the accuracy of claims. Narrative communication aims to provide reasonable depictions of individual experiences, judged on the plausibility of situations. Those are different success criteria, which is why selling can feel inappropriate and storytelling can feel like manipulation to a founder who has spent a decade in the first mode.

What is a two-layer narrative?

A decision layer that earns more diligence time, and a diligence layer that survives it.

Investors cannot process every technical detail with equal depth during an initial screen. They use prior patterns, category knowledge, perceived fluency, and the questions the first explanation generates. Jargon signals expertise to a specialist and increases cognitive load for a generalist when it is not connected to a business consequence.

Decision layer - customer
Pain, quantified in money, time or mission failure
Cost of inaction
Why now, with evidence rather than optimism
Product and the evidence it works
Business model and the financing milestone
Risk, named before it is asked about

Diligence layer - architecture
Benchmark protocol and validation data
IP position and freedom to operate
Regulatory pathway
Manufacturing and deployment
Unit economics detail
Technical limitations, stated plainly

This is not simplification for its own sake. It is information architecture. The core deck enables a decision about whether to spend diligence time; the appendix enables that diligence. Simplification is what gets you funded, partnered and scaled, and it is the thing founders resist hardest because years of research sit behind every slide.

Why does the evidence have to be encoded, not just included?

The relevant claim is not that investors are irrational or that storytelling overrides evidence. It is that evidence must be encoded in a form that lets the listener identify what it proves, what it does not prove, and which question should be tested next. A deck that leaves those three ambiguous fails regardless of the quality of the science underneath.

Market sizing is where this breaks most often. Around 55% of decks reviewed in 2024 carried inadequate market analysis, and bottom-up sizing is currently treated as roughly twice as credible as top-down. Published benchmarks for early-stage B2B companies sit at 1% of serviceable market in year one, 3% in year two and 5% in year three (MicroVentures, January 2026). Claiming 10% in year one without an operational plan raises immediate flags, and a serviceable-obtainable figure that implies $45 million of revenue while the model shows $20 million destroys credibility on the spot.

"It's like a Star Wars shooter" is one of the best pitch openings I have heard. The technology shared nothing with the film. The analogy created a slot in an overwhelmed investor's head for the innovation to occupy, and engaged everyone in the room who grew up on the franchise. An analogy is not dumbing down. It is addressing.

Which narrative techniques actually hold up?

Each technique has a failure condition and a measurement, and neither is usually discussed in pitch coaching.

1. Problem-first framing. Establishes relevance before mechanism. Fails when the problem is generic or unquantified. Measure: can the investor restate your customer and their pain afterwards.
2. Consequence framing. Makes urgency economically legible. Fails when the cost of inaction is inflated. Measure: the quality of the questions coming back.
3. Analogy. Gives a mental handle for unfamiliar technology. Fails when it implies the wrong mechanism. Measure: comprehension, and how often you must correct the technical picture afterwards.
4. Why-now evidence. Connects timing to external change. Fails when the trend is unrelated to a buying trigger. Measure: second-meeting conversion by narrative version.
5. Category naming. Helps investors place you. Fails when the category is vague or unsupported. Measure: whether investors describe you consistently to each other.
6. From-to transformation. Shows what changes for the customer. Fails when it ignores adoption and switching costs. Measure: customer validation and pilot scope.
7. Audience-specific versions. Reduces irrelevant detail. Fails when versions create inconsistent claims. Measure: conversion by investor type.
8. Technical appendix. Preserves credibility without overload. Fails when the appendix hides unresolved risk. Measure: diligence progression and question quality.
9. Objection pre-emption. Signals realism. Fails when it sounds defensive or over-scripted. Measure: time to resolution of the key diligence issue.
10. Rehearsal. Improves clarity under pressure. Fails when it polishes delivery without fixing the evidence. Measure: recorded comprehension and Q&A performance.

What does the from-to transformation look like in practice?

The pattern is visible in companies that have raised at scale. Relativity Space, valued at $4.2 billion, could describe wire-arc and powder-bed metal additive manufacturing systems for large monolithic aerospace structures in proprietary nickel and aluminium alloys. What they say instead is that they print rockets like software updates. Recursion Pharmaceuticals, at $2.13 billion, integrates high-throughput phenomics, transcriptomics, proteomics and ADME profiling into a unified data pipeline - or, as they put it, microscope-based AI for drug discovery. PsiQuantum, past $1.3 billion, builds photonic quantum computers using silicon photonics fabricated at existing semiconductor foundries: quantum chips built in the world's existing chip factories, at industrial scale.

Three golden rules produce those lines. Output, not process, because nobody cares how the technology works and everybody cares what they get out of it. Analogy first, to get through the cognitive barrier before the audience's attention closes. And the so-what test after every technical claim, repeated until you land on a currency figure, a minute saved or a risk avoided.

How do you measure whether the narrative worked?

Attention is not belief and belief is not investment, so measure comprehension and the next decision.

Deck analytics measure viewing behaviour. A slide viewed for longer may be confusing rather than compelling. A short deck may perform well because the investor already knew the company, not because a technique caused the outcome.

The sequential measures are stronger. Does the investor correctly restate the problem and the customer? Does the investor name the same why-now trigger you do? Does a second meeting happen with a relevant technical or commercial question? Does diligence surface fewer avoidable misunderstandings? Does the investor proceed, pass, or ask for a different evidence package? Each of those connects comprehension to a decision, which is what a single engagement statistic cannot do.

The same logic applies on the commercial side of the story. Enterprise buyers are lost to inaction more than to competitors: between 50% and 70% of qualified deals in complex B2B end in no decision at all. A narrative that does not make inaction expensive is competing with nothing, and losing.

Which pitch advice turns out to be folklore?

Five widely repeated claims have no traceable primary evidence, and two of them were in the first draft of this article.

1. A universal investor attention span measured in seconds or minutes. The circulating figures require primary investor-funnel datasets that are not publicly accessible. I used one in an earlier version of this piece. It came out.
2. A universal best deck length or slide order. No accessible dataset supports a fixed number for deep tech. Ten to twelve core slides plus a curated appendix is a working convention, not a finding.
3. The claim that passion alone drives investment. The frequently cited fMRI study on founder passion and investor interest could not be verified from accessible sources. The retrievable neuroscience literature concerns financial risk-taking and investment decision-making in general - for instance an ALE meta-analysis on investment decision-making, and work on insula activation and self-reported stock trading - not a verified founder-pitch experiment. Treat the specific percentages as unsupported until the original paper, sample, task and effect size are checked.
4. The claim that a category name creates a market by itself. Category creation is a business decision with marketing consequences, not a marketing act with business consequences.
5. The claim that narrative compensates for missing evidence. It does not. It determines whether the evidence you have is understood.

Deep tech pass rates, market-sizing failure rates and deck-screening funnel data would all require primary datasets or academic replication that are not currently available. If someone quotes you a number in one of those categories, ask for the denominator before you repeat it.

What should a founder do differently on Monday?

Write the narrative document before touching the deck, and test comprehension on a stranger.

The narrative document is six to eight pages and it is the source every other asset derives from - deck, outreach emails, the Q&A sheet, the forwardable summary. Build it in the order the investor needs: why the problem is inevitable, why now, why you, how the market maths hold up bottom-up, what proves the business model, and the ask.

Then write a jargon-free newspaper headline of under fifteen words before every investor meeting. If you cannot, the story is not ready. And name the discount yourself - investors apply a standard haircut to projections, and bracketing your own assumptions reads as self-awareness rather than weakness.

Frequently asked questions about deep tech investor narrative

Can we skip the narrative document and just fix the deck?

You can, and the deck will contradict your emails within a fortnight. The narrative document is the single source every other asset derives from, which is why fixing slides in isolation produces consistency problems that surface in diligence - a market figure in the deck that does not match the model, a why-now in the email that differs from the one on stage. The document takes a day to write and prevents weeks of that. Ask me about the narrative work.

How long should a deep tech deck be?

There is no evidenced universal answer, and anyone quoting one is repeating convention as if it were data. Structure matters more than count: a decision layer that earns diligence time, and a curated appendix that survives it. Ten to twelve core slides is a common working range in deep tech, with the technical detail moved into the appendix rather than deleted. What reliably fails is a single deck attempting both jobs at once.

What are the fifteen hardest investor questions?

Yours specifically, not a generic list. They are the objections your traction, market and team invite: timing risk, market-size credibility, moat durability, team commercial gaps, manufacturing scale-up, regulatory classification, and the discount investors apply to your projections. The exercise is writing your answers before an investor asks, in your own words, and marking which ones you cannot yet answer. That last list is your fundraising plan.

Our science is genuinely complex. Where does the technical detail go?

Into the diligence layer: a curated appendix with benchmark protocol, validation data, IP position, regulatory pathway, manufacturing plan and stated limitations. That is where your PhDs meet theirs. The failure mode is using the appendix to hide unresolved risk, which diligence finds anyway and reads as either disorganisation or concealment. Disclosing limitations there is the point rather than the cost.

Does rehearsal actually change fundraising outcomes?

It improves clarity under pressure, which is measurable through comprehension and Q&A performance. It does not fix missing evidence, and polished delivery over weak proof is a documented failure condition - a well-rehearsed pitch simply gets to the real objection faster. Record one rehearsal, then check whether a listener can restate your customer, your why-now and your ask without prompting.

Why did you delete statistics from this article?

Because they could not be traced to a primary source, and a number that fails in diligence damages more than the absence of a number. The investor deck-reading time figure and the fMRI passion percentages were both in the first draft. Removing them left the argument intact, which is itself informative: the techniques did not depend on those numbers. See why unverified figures get flagged.

Should the narrative differ for a generalist versus a deep tech investor?

Yes, in emphasis rather than in facts. A generalist needs the commercial consequence first and the mechanism second; a specialist fund with technical partners will move to the diligence layer quickly and judge you on the protocol. Maintaining audience-specific versions is a recognised technique, and its failure condition is versions that make inconsistent claims. Keep one set of facts, two orders of presentation.

How does this differ from a SaaS pitch?

Two structural asymmetries change the emphasis. Pivot cost: a SaaS company refits its solution to a moving problem in months, while a deep tech company needs eighteen to thirty-six months, and regulated biotech longer - so the narrative must show the application was validated before launch rather than promise agility. And education: SaaS sells a better way to do a known thing, deep tech sells a new way to do a new thing, so the story teaches before it sells.

Six weeks out is the right time to fix the story. A 30-minute call on where the raise stands, what exists, and what is missing. If the narrative is not the problem, I will say which part is.

Book a 30-minute call · Read the traction article

What changed in this article (7 September 2026)

Removed the investor deck-reading time figure as unverifiable

Removed the fMRI passion percentages and added what the accessible neuroscience literature does and does not cover

Added the two-layer narrative model and a failure condition plus measurement for each technique

Added the from-to examples from Relativity Space, Recursion and PsiQuantum