Pitch and Demo: Let Them Say the Problem, Then Let Them See the Result
Topic: pitching and live demos·4 moves · 1 sentence
The most complete pitch usually loses to the shortest one — they walk out with a single thing, and the only question is whether you chose it or they grabbed it at random.
When you fail to convince someone, it's rarely because you said too little — it's usually the order: lead with the solution and their job becomes hole-hunting; describe their problem until they nod and their job becomes adding detail. Same material, different order, different role for them. And a harder one to swallow: after a twenty-minute demo, the only things they can repeat the next day are one idea and the last image on the screen — what goes in those two slots is the only part you genuinely control.
MOVE 01
Problem first, then the fix
Problem first, then the fix
diagnose firstcuriosity gapframing
The principle in one line
Before you present the fix, describe their problem until they nod or correct you. A solution with no problem behind it can only be graded, and grading means hunting for holes; get the problem right and they start adding detail on their own — which is the moment you end up on the same side of the table.
In their words
"A problem well stated is a problem half solved."— Charles F. Kettering (head of research at General Motors); long attributed to him, so treat it as folk wisdom rather than a citation
Same twenty minutes, two orders
How it plays out
Setup: you sell staff-scheduling software and you're meeting the operations director of a restaurant chain. You have twenty minutes.
✗ Starting with yourself
"Let me introduce us briefly — five years in business, three hundred customers, and the platform has five modules…" —— For three minutes they're waiting to hear how this concerns them; for fifteen more they're grading each module. By the time you name a price they hold a pile of features and not one problem of their own, so the cheapest possible answer is "send me a deck."
✓ Starting with their Wednesday afternoon
"Let me start with what I think is going on — stop me if I'm off. Every Wednesday afternoon you spend two or three hours building next week's rota; someone swaps a shift and it costs you seven phone calls; at month end you reconcile hours against the clock-in data. But I'd guess the expensive part isn't those two hours — it's the day you're overstaffed by two people. Is that right?" —— They'll probably answer "worse than that — there's also finding cover for sick days." The second they start adding, you stop being a vendor and become the person they brought in to look at the problem.
Why it works
Why it works
Mechanism: naming the problem first leaves an open gap, and the tension holds attention — people want the closure. It also switches their role from judge to co-diagnostician. Leading with the solution does the opposite: a fix with no problem behind it can only be scored.
Psychology · curiosity is a gap, not an interest (well-supported review): Loewenstein's (1994, Psychological Bulletin) information-gap theory holds that once people notice a gap between what they know and what they want to know, it acts like a deprivation state. State the problem clearly and the fix suddenly has an audience.
Neuroscience · curiosity recruits reward circuitry and boosts memory (hard evidence): Kang et al. (2009, Psychological Science) found that high curiosity increased activity in the caudate and inferior frontal gyrus (part of the reward circuit) and predicted better later recall of answers; Gruber, Gelman & Ranganath (2014, Neuron) showed that curious states co-activate the midbrain SN/VTA dopaminergic region and the hippocampus, with even incidental material presented at the time remembered better. Translated: the two minutes they're curious are the window in which anything you say sticks. (cross-ref neuroscience on reward and memory, psychology on curiosity)
Practitioner framework · useful, but don't believe its neuroscience (labelled honestly): Oren Klaff's Pitch Anything (2011) argues a pitch is first screened for threat and novelty and only then judged on content — that's field experience, not peer-reviewed evidence, and his "crocodile brain" story is anatomically wrong (the triune-brain model was discarded long ago; see Cesario, Johnson & Eisthen 2020, Current Directions in Psychological Science). But his operational advice — short, concrete, novel, non-threatening, end by asking for a decision — points the same way as the real evidence above. Follow the practice; don't adopt the biology.
Boundary: "problem first" is not manufacturing panic. The problem you name must be one they recognise and could restate in their own words; inflating a non-existent problem into a crisis works this quarter and destroys the account. Fear-appeal research agrees: an appeal only works when it comes with a specific action they believe they can take, otherwise people avoid the whole subject. Guessing the pain wrong is fine too: the correction they give you is worth more than a lucky guess, which is why every version of this ends with "stop me if I'm off." The quantifying questions from the pain-digging issue slot straight into the "cost in money" step.
Useful phrasings
Say it like this
"Let me start with what I think is going on — stop me if I'm off." —— the best opening line there is.
"Is that roughly right, or am I missing something?" —— invites confirmation or correction.
Sell the problem before you sell the fix.
MOVE 02
Less is more: one sentence survives, so place it deliberately
One takeaway, placed on purpose
working memoryprimacy and recencyredundancy effect
The principle in one line
They leave with three or four slots, and the opening and the close claim the best two. So pick the one sentence first — their outcome, not your feature — say it at the start and again at the end, and make everything else give way, including the words on the slide that compete with your mouth.
In their words
"Focusing is about saying no."— Steve Jobs, WWDC 1997 Q&A session
The slots they actually have
How it plays out
Setup: the same twenty minutes, and you've prepared 26 slides.
✗ Pouring out what you prepared
Five bullets a slide, and you read them aloud. —— Two bad things happen at once: they read while you narrate the same words, one channel gets loaded twice and comprehension drops; and nowhere in 26 slides is there a designated sentence, so they grab something at random — usually the last thing they saw, a screenshot of the settings page.
✓ One sentence throughout, slides get out of the way
First line of the meeting: "There's only one thing today — you never build a Wednesday rota again." Every section hangs back on it ("this step removes the shift-swap calls, the worst part of those two hours"). Slides hold a headline and one diagram: what's written is what you don't intend to say. The close: "Back to where we started — no more Wednesday rota." —— The 26 slides' content isn't lost; it becomes the one-page annex you send afterwards.
Why it works
Why it works
Mechanism: how much someone can hold at once is counter-intuitively small, and position decides which pieces survive. Worse, you can jam the channel yourself — when your narration and the screen carry the same words, the two inputs interfere, and the fuller you make it the less they understand.
Cognitive psychology · about four chunks at once (well-supported; the number is debated): Cowan (2001, Behavioral and Brain Sciences) reviewed a large body of work and put the capacity near 4±1 once rehearsal and chunking are ruled out; Miller's famous "7±2" (1956, Psychological Review) is the ancestor of this finding but was later revised. Honest label: the exact number varies by task; what's solid is that it's very small.
Memory research · primacy and recency (classic and highly replicable): Murdock (1962, JEP) produced the standard U-shaped serial position curve in free recall — items at the start and the end are recalled far better than the middle. Putting your key sentence in the middle is volunteering to lose it.
Hard evidence · narrating text they're also reading makes comprehension worse (the redundancy effect): Kalyuga, Chandler & Sweller (1999, Applied Cognitive Psychology) and Mayer & Moreno (2003, Educational Psychologist) repeatedly found that presenting identical on-screen text and narration hurts understanding, because both compete for the same processing channel — one of the rare "give less, learn more" results in education research. The action is unambiguous: put on the slide what you don't plan to say.
Boundary: you're cutting what you say this time, not what you prepare. In complex deals they have to go and convince other people with your material (the ones the decision-chain issue is about), so write the detail into a one-page annex and send it after — not spreading it in the room isn't the same as withholding it. Answer detail questions from technical audiences, and answer them better than they expect: less-is-more governs what you volunteer, not how deep you go when asked. The other failure mode is picking the wrong sentence: if it describes your feature instead of their outcome, saying it ten times won't help.
Do this
Write the one sentence on paper; it's the only thing allowed to be said twice (open and close, word for word)
For each slide ask "if I delete this, does the sentence still stand?" — if yes, delete it
Leave a headline and a diagram on screen; move your script into the speaker notes
Turn the cut material into a one-page annex and say "detail's on one page, I'll send it after"
Cut the weakest slide on purpose
Useful phrasings
Say it like this
"If you take one thing away from today, take this one." —— designate the sentence.
"The details are in a one-pager — I'll send it after." —— cut cleanly without looking unprepared.
Don't read your slides to me; don't make me read them either.
MOVE 03
Demo rhythm: result first, pauses inside, the peak at the end
Result first, pauses inside, peak at the end
peak-end ruleinterpolated questionsno feature tour
The principle in one line
Don't run the demo in your product's order; run it as result → how → their hands on it. Show the finished thing inside the first minute, stop every few minutes to collect one response, and land the ending on the strongest moment — because what they keep afterwards is the peak and the last second, not how long you demoed.
In their words
"We don't choose between experiences, we choose between memories of experiences."— Daniel Kahneman, TED 2010 talk "The riddle of experience vs. memory"; the same argument runs through Thinking, Fast and Slow
A twenty-minute demo, timed
How it plays out
Setup: same scheduling product, screen shared, twenty minutes on the clock.
✗ Demoing in product order
"I'll log in first — here's permission setup, three levels of access…" —— Twenty minutes later you're on the settings page. The session has no peak, the last image is a screen full of empty fields, and in memory the product now equals "a lot of setup." The more thoroughly you demo, the firmer that impression gets.
✓ Result to open, peak to close
"Result first: this is last week's rota for your Wangjing branch, generated in five seconds, with the two overstaffed slots in red. (two-second pause) How long does that sheet take you today?… Two hours plus. Right — let me rewind and show you where those five seconds come from." The close returns to the sheet: "That's the two hours back in your Wednesday afternoon. Two options for the next step: we run your three branches' real data, or start with Wangjing only?" —— The peak lands in the first minute and is stepped on again at the end; everything in between hangs on that sheet.
Why it works
Why it works
Mechanism: memory of an episode isn't the sum of its minutes — it's roughly the average of the most intense moment and the final one, with duration barely entering the account. And people who only watch drift off, while being asked for a response every few minutes pulls attention back.
Hard evidence · memory runs on the peak and the ending, not the length: in Kahneman, Fredrickson, Schreiber & Redelmeier (1993, Psychological Science), participants preferred to repeat the longer cold-water trial that ended more comfortably; Redelmeier & Kahneman (1996, Pain) replicated it during real colonoscopies — overall pain ratings tracked the peak and the final moments while duration barely mattered ("duration neglect"). So don't end on the settings page, and don't grind out 40 minutes to look diligent.
Hard evidence · one response mid-way substantially reduces mind-wandering: Szpunar, Khan & Schacter (2013, PNAS) segmented online lectures and inserted a short test after each segment; mind-wandering dropped, note-taking rose, and final test scores improved. The sales version is one calibrating question every four or five minutes — not politeness, but re-attaching their attention.
Honest label · "attention only lasts ten minutes" is folklore, not a finding: Wilson & Korn (2007, Teaching of Psychology) went back through this endlessly quoted claim and judged the original evidence weak. So don't treat "change the act every ten minutes" as law; what's actually supported is making them respond, not showmanship.
Boundary: leading with the result assumes the result is real. Running their own data live is the most convincing thing you can do and the easiest to crash — test it in your own environment first and keep an offline version ready. Debugging in front of them is the worst damage this move can cause, because it destroys the peak and the ending at once. Also, don't read "peak" as showing off: their peak is the second their own headache disappears, not your animation; and never invent a promise to get a good ending — the close should be something you already demoed, stepped on one more time.
Useful phrasings
Say it like this
"Let me show you the end result first, then we'll rewind." —— open on the result.
"Before I go on — how does that compare with how you do it today?" —— the mid-way pause.
End on the moment that mattered, not on the settings page.
MOVE 04
Get them involved: one thing they click themselves beats ten you demo
Put their hands on it
generation effectself-referenceownership
The principle in one line
The same thing, produced once by their own hands, sticks harder and feels far more like theirs than anything you demonstrate. So: use their real data, hand over the mouse once (two fields is plenty), and let them make one very small choice in the room.
In their words
"People are generally better persuaded by the reasons which they have themselves discovered than by those which have come into the mind of others."— Blaise Pascal, Pensées (1670)
The involvement ladder: how far one demo can climb
How it plays out
Setup: minute 14 of the demo. They've been watching and haven't touched the keyboard.
✗ You drive throughout, then invite them
"That's roughly it — would you like to sign up for a trial?" —— The jump from watching to starting from scratch alone is far too big, and the whole cost sits on their side. The cheapest answer is always "sure, send me a link and I'll look when I get a chance," which in practice means this is over.
✓ Thirty seconds, two numbers to type
"Here, you drive — just two numbers: how many people did you roster at this branch last Wednesday? And on the lunch shift? (they type) Good, hit enter. See those two red cells? That's the two extra people last week." Then a small choice: "Do you want to look at labour cost first, or sick-day cover?" Before they leave: "I'll export the sheet you just built and send it over." —— They'll forward it to the store manager and to finance, which is exactly the thing the decision-chain issue says is hardest to get: material that argues for you when you aren't in the room.
Why it works
Why it works
Mechanism: material you generate yourself goes through deeper encoding than material you receive; anything that attaches to "me" gets special treatment; and things you handled yourself get valued higher automatically. Stack the three and the owner of the plan in their head quietly shifts from you to them.
Hard evidence · self-generated material is remembered better (the generation effect, highly replicable): Slamecka & Graf (1978, JEP: Human Learning and Memory) had participants complete cue words themselves rather than simply read them, and recall improved substantially. The result they produced by pressing the key and the identical result you showed them are not the same object in their memory.
Hard evidence · information about "me" is processed differently (with a brain mechanism): Rogers, Kuiper & Kirker (1977, JPSP) established the self-reference effect — material judged for "does this describe me?" is remembered best; Kelley et al. (2002, Journal of Cognitive Neuroscience) found self-referential judgements selectively engage the medial prefrontal cortex. So running their company's real data isn't just courtesy; it hooks the content onto their self-representation. (cross-ref neuroscience on self-reference, psychology on self-related memory)
Consumer behaviour · handling raises valuation (reliable, moderate size): Norton, Mochon & Ariely's (2012, Journal of Consumer Psychology) "IKEA effect" — people value what they assembled themselves more highly; Peck & Shu (2009, Journal of Consumer Research) found that mere touch raised felt ownership and willingness to pay. Honest label: these effects are real but not large — they can push "decent" into "I want it," not "we don't need this" into a sale.
Boundary: this move sits closest to manipulation, and the line is clear — hands-on exists so they can judge for themselves whether this is any use to them. If they touch it and discover it doesn't fit, that's the most valuable output of the whole demo (exactly the time-saving the qualifying issue is about). Using the time they've already invested to block their exit ("you've spent two hours on this") turns the IKEA effect into a noose: you close today and collect refunds, bad reviews and one person who never takes your call again. The second boundary is friction: two fields is fine, but making them export data or configure permissions on the spot becomes work — the denominator from the irresistible-offer issue applies here too.
Do this
Get one piece of their real data before the meeting (a sheet, an order number, a real piece of copy)
Design a thirty-second hands-on moment with at most two input fields
Prepare a small either-or choice so they make one decision in the room (a tiny one)
Send them the exact output they produced, so they have something to forward
If they say "this doesn't look right for us" after touching it, thank them, then ask which step doesn't fit — don't rush to rescue it
Useful phrasings
Say it like this
"Here, you drive — just type these two numbers." —— hand over the mouse.
"I'll send you the version you just built." —— give them something forwardable.
What they build, they believe.
Your Day 17 Action
Take the next pitch or demo you owe someone and write four lines before you touch a slide.
1 (10 minutes): write their problem in their own words, ending with "stop me if I'm off"; then one line of the cost you'll ask them to confirm (in money or hours).
2: pick the one sentence — their outcome, not your feature. Write it on paper, say it at the open and the close, word for word. Then move everything you planned to read into the speaker notes and delete the weakest slide.
3: set the two endpoints of the demo: which finished result they see in minute one, and which moment you land on at the end (it must be something already demoed). Insert two calibrating questions, around minute 5 and minute 12.
4: design a thirty-second "you drive" moment — two input fields at most, using their real data — and close by asking for an either-or decision.
Boundary reminder: all four rest on the pain being real and the result being real. Inflating a non-existent problem into a crisis, demoing a result that doesn't actually run, using their invested time to block their exit — each raises today's close rate, and each is collected back with interest on delivery day.
Think It Through
1. They give me fifteen minutes, or open with "just tell me what you do." Still lead with the problem?
Yes, but compress it to twenty seconds and deliver it as a statement, not a question. "Just tell me what you do" usually means "don't waste my time," not "skip the diagnosis" — spending two minutes asking background questions annoys them, while one accurate summary doesn't, because the summary is itself information. So: one sentence of the problem as you understand it, plus "is that right?" A nod gives you the frame; a correction gives you something worth more than a lucky guess. A fifteen-minute cut looks like this: 20 seconds confirming the problem → 3 minutes of one result demo → 5 minutes on how it works (with one calibrating question) → 2 minutes of their hands on it → the rest asking for a decision. There's also an opening that suits impatient people better: a verifiable result from someone like them ("three chains your size cut that weekly two hours down to about ten minutes"). That's still the pain, just told from the result end — and it plants a reference point on the way past.
2. If I say that little, won't I look unprepared or unprofessional?
That worry usually comes from treating "prepared" and "poured out" as the same thing. Perceived expertise comes from two places: one, your description of their problem is sharper than the way they'd have put it themselves (which requires preparing far more than you say); two, when they ask for detail, you answer fast and precisely. So the right posture is prepare tenfold, present one thing. A concrete insurance policy: turn the cut material into a one-page annex and say in the room, "I put the detail on one page, I'll send it after" — that single line proves both that you have it and that you know what deserved airtime, and the second is the scarce part. Conversely, if saying less genuinely leaves you sounding thin, it's usually not the volume, it's the wrong sentence: it describes your feature rather than their outcome, so it lands like a slogan. The test is simple — can they answer "what is that worth to me?" in their head after hearing it.
3. During the demo they keep picking at features, or ask for one we don't have. What now?
First tell a buying signal apart from evaluation reflex, then three steps. One, never commit to a roadmap on the spot — the most expensive mistake there is; "next quarter" gets collected at full price on delivery. Two, translate the feature question back into a result question: "which step were you planning to use that for?" Two common outcomes: what they want is something you can already do another way, or one question reveals they were just reading off a comparison checklist. Three, if you genuinely don't have it and it genuinely matters, say so plainly, and spell out what it costs them and what the workaround is — the credibility you buy by admitting a gap beats a vague promise, which you'd also have to repay later. One last diagnosis: if a whole demo circles the feature list, "problem first" wasn't done well enough — they haven't accepted the framing of the problem, so a feature checklist is the only ruler they have. The most effective move then isn't to keep demoing; it's to step back and ask about today: "how do you handle this step at the moment?"