Every interface is a claim about how a mind works.
How Human Psychology Decides Whether Your Software Works.
scroll · het boek gaat open
de inzet
A form with twelve fields claims a person can comfortably track twelve things.
A feature announcement in a colorful box claims people read pages the way you read code review comments: thoroughly, top to bottom.
A pricing page with two plans claims prices are evaluated on their merits, in a vacuum, by a calm evaluator with time.
A confirmation dialog claims people read confirmation dialogs.
05 · Deze pagina claimt dat je alle vier gelezen hebt.
Most of these claims are false.
The bug isn't in your code. It's in your model of the person running it.
Most of these claims are false, and they are false in consistent, measurable, well-documented ways. Psychologists have spent a century mapping how perception, memory, decision-making, and motivation actually behave, and the map disagrees with the folk model most software is built on. The folk model says users see what's on screen, remember what they were shown, weigh options rationally, and act on their intentions. The map says users see a narrow trail plus guesses, hold about four things at once, judge everything relative to whatever stood next to it, and do what the default suggests.
deel 1 · de specificatie01
The Machine You're Building For
The recording is eleven minutes long. Nobody at Lintel makes it past minute six without wincing.
01working memory, capacity of about four
Working memory is the mind's scratchpad: the handful of items you can hold in awareness right now, this second, while doing something with them.
02attention runs single-file
Deliberate attention processes one thing at a time.
03two speeds, and the fast one is driving
Thinking runs at two speeds.
04perception guesses
The eye is not a camera and the visual system is not a recorder.
Sterk
de wandeling · twaalf spreads
Door het boek
01 / 12
The Bug Is Not in Your Code
02 / 121 · The Spec
The Machine You're Building For
The recording is eleven minutes long. Nobody at Lintel makes it past minute six without wincing.
03 / 122 · The Path
Getting In
For two quarters, Lintel's most requested feature was one they had already shipped.
04 / 122 · The Path
Holding On
Lintel's onboarding funnel had a completion rate of 11 percent, and the team could prove it wasn't their fault.
05 / 122 · The Path
Deciding
Two founders, ex-security engineers, built a genuinely good secrets-scanning tool and priced it at nine dollars a month.
06 / 122 · The Path
Other People
Bramble's landing page had social proof the way a hotel lobby has art: professionally sourced, expensively framed, and invisible.
07 / 122 · The Path
Feeling
In the same ugly quarter, Lintel and its closest competitor, Slatewire, both went down.
08 / 122 · The Path
Saying It
The rebrand cost Lintel a hundred and forty thousand dollars, and you could feel the money in every sentence.
09 / 122 · The Path
Coming Back
Stride is a habit app: workouts, stretching, the occasional guilt-free rest day, streaks to bind it all together.
10 / 123 · The Discipline
The Aiming Discipline
The experiment was called `cancel_flow_v3`, and it won.
11 / 123 · The Discipline
The Folklore Graveyard
The deck was called "Design Principles Q3," and slide nine was a massacre that nobody in the room could see.
12 / 12
A Year Later
een fragment · hoofdstuk 7
Saying It
The rebrand cost Lintel a hundred and forty thousand dollars, and you could feel the money in every sentence. The old homepage had said "Match your invoices to your bank statements automatically." The agency's new one said "Empower your finance workflow with intelligent reconciliation, reimagined."
Demo requests fell 30 percent in six weeks. The team's diagnosis followed the money: "The new positioning isn't resonating with enterprise. Recommend a messaging workshop." A second workshop, to fix the first workshop.
Meanwhile, over at Stride, a smaller and stranger fire was burning. The habit app's new paywall had been written by someone who'd clearly read about objection handling. Under the price it said: "No hidden fees. No long-term commitment. Don't worry about cancellation hassles." Reassurance, wall to wall. Within a month, support was fielding a new genre of ticket, arriving daily: "What are the hidden fees?" "How hard is it actually to cancel?" Users were asking, with visible suspicion, about problems the copy had explicitly denied. The copywriter was baffled. She had answered those questions in advance, in bold, on the page.
Reading doesn't work like transmission. It works like rendering.
Copy is source, and the reader is the renderer.
deel 3 · de discipline10
The Folklore Graveyard
Four load-bearing claims. Zero survivors. And the roadmap was already funded.
Achttien overtuigingen die als wetenschap rondgaan. Open er een en lees wat er gebeurde toen iemand het narekende.
Willpower is a muscle that depletes
de overtuiging
self-control is a fuel tank; decisions drain it; users (and you) run dry by evening, so schedule big asks early.
bij hertest
the foundational "ego depletion" experiments went through large multi-lab replications and produced effects near zero. The famous supporting anecdote (judges granting parole less often before lunch) has strong alternative explanations in how cases were ordered.
in plaats daarvan
what survives is structural, and it's enough: across long sequences of choices, people default, defer, and abandon more as the sequence grinds on. Put the decisions that matter early, collapse trivial ones into good defaults, and never write "decision fatigue" in a strategy doc as if it were a fuel gauge.
People are left-brained or right-brained
de overtuiging
analytical users and creative users are different neural types; design and market to each.
bij hertest
brain imaging on over a thousand people found no such trait; specific functions lateralize, personalities don't.
in plaats daarvan
if you need to segment users, segment on observed behavior and expertise, which actually predict things.
People have learning styles (visual, auditory, kinesthetic)
de overtuiging
match content format to each learner's style and learning improves.
bij hertest
the styles exist as preferences; the matching benefit is the claim, and controlled tests keep failing to find it.
in plaats daarvan
match format to the content (diagrams for spatial structure, narration synced to demonstration), and use what does replicate at scale: spaced exposure, retrieval practice ("now you try"), and pictures paired with words.
Communication is 55% body language, 38% tone, 7% words
de overtuiging
on your demo call, the words barely matter.
bij hertest
nothing failed, because the claim was never made. Mehrabian's 1967 experiments measured something tiny: judgments of a speaker's feelings when a single spoken word conflicted with tone or face. He spent his career objecting to the generalization.
in plaats daarvan
delivery affects trust and liking; content carries the information. Prepare both; believe no percentages.
Incompetent people think they're experts (the pop Dunning-Kruger)
de overtuiging
the least skilled are the most confident, per the famous graph.
bij hertest
the real finding is milder: people are miscalibrated about themselves, and much of the dramatic graph shape reproduces even in random data, a statistical artifact. The "confident idiot" reading overstates it badly.
in plaats daarvan
the design lesson survives in useful form: never trust self-reported expertise. Don't ask "are you a beginner?"; watch behavior and adapt to that.
Eyes scan pages in a Z-pattern
de overtuiging
layout your hero, benefits, and CTA along a Z, because that's how eyes travel.
bij hertest
there was nothing to retest; the Z was a diagram in design blogs, not a finding. Actual eye-tracking shows content-driven scanning: the F-shaped pattern on unstructured text walls (a symptom of users giving up, per the researchers who named it) and heading-to-heading hops on structured pages.
in plaats daarvan
Chapter 2's program: build real hierarchy, front-load headings and sentences, put the payload where sampling lands, and make the primary action the one thing that survives a squint.
Blue means trust
de overtuiging
blue induces trust; that's why banks use it; paint your brand accordingly.
bij hertest
no wavelength-to-trust wiring survives scrutiny. Blue is the world's most broadly liked color (a safe choice, and that part is real preference data), and banks look trustworthy in blue because banks have looked blue for a hundred years. The trust lives in the category convention, in looking like what you claim to be.
in plaats daarvan
choose palettes for category fit, distinctiveness, and contrast. If you're a bank, blue is fine, for the honest reason: costume, not chemistry.
There's a highest-converting button color
de overtuiging
the famous tests proved red beats green (or orange beats everything); pick the winning hue.
bij hertest
reread the famous tests and they were never about hue. The winning button was the one that contrasted with everything around it, the lone odd-one-out that Chapter 2 called pop-out. Red beat green on a green-tinted page.
in plaats daarvan
pick any accent that maximally contrasts with your palette, use it exactly once per screen, keep it consistent across the funnel, and never debate hue in a meeting again.
Menus should have 7±2 items
de overtuiging
Miller proved working memory holds seven things, so visible options must be capped at seven.
bij hertest
two failures stacked. The modern capacity estimate is about four chunks, and, more fatally, the limit governs what must be held in the head, never what's visible on screen. A twelve-item menu the user can see costs almost no memory; the screen is doing the remembering. Hiding options behind "More" to satisfy the quota measurably hurts findability.
in plaats daarvan
spend the discipline where the limit is real: nothing important carried across screens, rules visible while typing, codes copyable. Organize long menus with grouping and order; don't amputate them.
Hard-to-read fonts improve memory (Sans Forgetica and friends)
de overtuiging
disfluency forces deeper processing, so ugly type makes content stick.
bij hertest
the original small finding failed to scale, and the purpose-built font showed no memory benefit in controlled tests. Meanwhile the opposite effect is among the sturdiest in the book: fluent, easy-to-read presentation makes content feel truer and tasks feel more doable.
in plaats daarvan
maximize readability everywhere you want belief or action. If you want retention, use the real levers: spacing, retrieval, self-relevance.
Dyslexia fonts help dyslexic readers
de overtuiging
specially shaped letterforms ease dyslexic reading; adopting one is an accessibility win.
bij hertest
repeatedly, no reliable benefit for the fonts themselves.
in plaats daarvan
what measurably helps is unglamorous: larger sizes, more letter and word spacing, shorter lines, ragged-right margins, strong contrast, and user control over all of it. Helps dyslexic readers, helps everyone else too.
Respond in 400 milliseconds (the Doherty threshold)
de overtuiging
a scientifically established productivity cliff at 400ms system response.
bij hertest
the source is a 1982 IBM business paper, not psychology, and the number was never a cognitive constant.
in plaats daarvan
the real, older structure: about a tenth of a second feels instant, about one second preserves the flow of thought, and about ten seconds ends the task's hold on attention. Engineer for the tier you can afford, and past it, manage the feeling of time with Chapter 6's tools: narrate, bound, occupy, and never stall at 99 percent.
More choices always kill conversion (the jam study)
de overtuiging
the 24-jam table sold less than the 6-jam table; therefore always cut options.
bij hertest
pooling dozens of studies produced an average effect near zero; assortment size cuts both ways, moderated by comparability, defaults, ordering, and the chooser's clarity about their own preferences.
in plaats daarvan
Chapter 3's rule: don't cut, aid. Recommended default, aligned comparison, honest sorting, named tiers. Cut options for strategy reasons if you like; just don't claim the psychology mandates it.
Subtle primes drive behavior (money pictures make people selfish, and cousins)
de overtuiging
ambient cues (imagery, words, colors) meaningfully steer downstream behavior, so sprinkle luxury cues and watch spending rise.
bij hertest
this family was the replication crisis's ground zero. The famous social-priming results (words about aging slowing people's walk, money imagery reshaping cooperation) failed direct replications, badly. What survives is the boring kind: semantic priming, where "doctor" speeds recognition of "nurse" by milliseconds. Real, tiny, and not a growth strategy.
in plaats daarvan
if you want context to sell, use context that informs: category-consistent imagery, coherent metaphors, congruent tone. Expect it to clarify, never to hypnotize.
Power posing changes your hormones and your outcomes
de overtuiging
two minutes of expansive posture before the pitch raises testosterone, lowers cortisol, wins deals.
bij hertest
the hormonal and performance claims collapsed in larger studies, and the paper's first author published a statement disavowing the effect. A small self-reported feeling of confidence may survive; nothing downstream does.
in plaats daarvan
the pre-demo intervention with actual evidence is preparation: rehearse the opening, pre-answer the top objection (Chapter 7's two-sided credibility), and get the demo environment boring and reliable.
Habits form in 21 days
de overtuiging
three weeks of streaks and the behavior becomes automatic; design trials and challenges to match.
bij hertest
the source was never research; it was a plastic surgeon's 1960 reflection on patients adjusting to their changed faces. When researchers actually measured habit formation, the median was about 66 days, ranging roughly 18 to 254.
in plaats daarvan
Chapter 8's program: anchor the behavior to an existing daily cue, use the "after I X, I will Y" sentence that reliably doubles follow-through, and keep scaffolding up for months, not weeks. Price your patience accordingly.
Your phone's mere presence drains your brain
de overtuiging
a silent, face-down phone on the desk measurably taxes working memory; cite the study, ban the phones.
bij hertest
the direct replications have mostly come back null. The claim outran its evidence years ago and kept getting cited because it's deliciously tellable.
in plaats daarvan
the attention costs that are solidly real: switching itself is expensive, interruptions leave residue on the next task, and notifications exploit reflexes you can't opt out of. Argue for focus features from those; they don't need the folklore's help.
It takes 23 minutes to refocus after an interruption
de overtuiging
every ping costs 23 minutes of productivity; multiply by pings; present the terrifying total.
bij hertest
the number is real but the claim is a misquote. The field research found people took that long to return to the interrupted working sphere, usually via a detour through other work, not that cognition lay in ruins for 23 minutes. Actual resumption lags run seconds to minutes, worst mid-task.
in plaats daarvan
the honest and still-damning version: interruptions cost real time and errors and their costs concentrate mid-task, so batch what can be batched and time what can't for natural breakpoints. Chapter 8's budget, again.
Zwak
Dat is de enige vraag.
epiloog · een jaar later
Lintel is geen getransformeerd bedrijf. Zo werkt het niet.
Lintel, twelve months on, is not a transformed company, because that isn't how this works. They still ship bugs. Their dashboard still has one chart nobody can explain. But four things are different, and all four are small. New-feature announcements go into the product's own furniture now, in the flow, in the house typography; nobody has proposed a violet box with a sparkle icon in three quarters. Onboarding asks two questions instead of nine, defaults the rest, and shows a new user their own matched invoices inside four minutes; completion sits at 61 percent, and the planning doc that once recommended shipping a salaried human with every seat has been quietly archived. The word "workspace" appears exactly once in the interface, because the glossary is enforced in code review now, like any other lint rule. And when the founder rewatches usability recordings (she still does, one per release, a habit that costs an hour and has outperformed every analytics contract they've signed), the write-ups no longer contain the phrase "not our target user." The last one ended with a different sentence: "She hesitated at step two. Find out what we did."
Je zag hem. Banner blindness is een gemiddelde, geen wet. (fig. 02)
de gewoonte
The audit reflex becomes a team habit.
het boek
De psychologie achter software die werkt.
Tien hoofdstukken, elke claim gelabeld op bewijskracht. Het volledige boek staat hieronder klaar.
Roy Heilbron · Elke claim in het boek draagt een label voor bewijskracht: sterk, gemengd of zwak.
