Perception-First Design™

The Bouncer

I figured out my design methodology in a nightclub.

I was head of security at a 300-capacity venue. Staff of 15. Every night at 1:30 AM, same problem: lights slam on, DJ announces we’re closed, and I’m barking at 300 drunk people to get out. They hated it. I hated it. Fights happened. Stragglers lingered. The whole thing was hostile.

So I tried something different. At 1:10, I walked table by table. “Hey, last call’s coming up. Get any drinks you want. Thanks for coming out tonight.” That’s it. No urgency. No authority voice. Just information, a suggestion, and respect.

At 1:30, when the DJ made the announcement and the lights went up, half the venue was already walking out the door. No barking. No fights. And bar sales went up 30% in those last 20 minutes.

I didn’t know what to call that. I just knew it worked, and I knew why it worked. The old way treated people like obstacles. My way treated them like people with their own goals who’d cooperate if you gave them the right information at the right time with the right tone.

That’s Perception-First Design before I had words for it. Everything I’ve done in the 15 years since; a 233% revenue turnaround, Costco pallets and Walmart shelves and Disney licensing, improv theaters and smart home brands. It all comes from that same instinct. Design for how people actually perceive and process the world, not how you wish they did.

The Thesis

The design industry has spent 15 years saying the same thing: users have limited attention, so don’t waste it. Lower friction. Reduce steps. Don’t make them think.

My version is different: users don’t think. Not until you make them.

Nobody is thinking. They’re on autopilot. The same way you catch a ball without calculating trajectory, or walk without planning each step; that’s how people interact with websites, products, brands, everything. Autopilot. Energy conservation. The brain handles the vast majority of processing unconsciously. Kahneman (2011) calls this System 1: the fast, automatic mode that runs most of cognitive life. System 2, slow and deliberate and conscious, only activates when something demands attention.

Contemporary cognitive science calls this predictive processing (Clark, 2013; Friston, 2010). The brain is constantly generating predictions about what it expects to encounter. When reality matches prediction: autopilot. When reality violates prediction: attention fires, System 2 kicks in. That prediction error is what I call the activation point.

The design challenge isn’t “don’t waste their limited attention.” It’s that effortful attention is dormant by default. The real question is: when you do make them stop and actually think, when you generate a prediction error that activates them from autopilot, what do they think about? How do they feel? And what do they do next?

That’s what I design for. Not the funnel. The activation.

Steve Krug wrote Don’t Make Me Think. Great book. It gave me the vocabulary to talk to stakeholders early in my career. His scope was usability, and he stayed in it deliberately. But nobody’s thinking in the first place. Design for that bias, the bias of making someone stop and engage, instead of funneling them to an action with an affordance of attention they never gave you.

Why I See This

Two things about my brain that I didn’t choose but that turned out to be professional advantages.

I’m autistic. After my formal ASD diagnosis, my partner pointed out something I’d been missing: “Do you realize you’re not picking up on me not being interested at all in what you’re talking about, and you seem to be getting frustrated?” Around 2012, a friend I met at a Chicago meetup about emotions in video games suggested I try improv. I went in already analytical. I had impaired social instincts, not absent ones. I didn’t need to learn what a smile means. I had to learn what eye contact can mean, what a pause signals, what “said vs. unsaid” reveals. I learned it like you’d learn an instrument or ballet: theory, practice, application loops, over and over.

That means I can see and articulate the unconscious patterns that most designers apply intuitively but can’t explain. People interact with websites the same way they interact with other humans. We form first impressions on products the same way we do when someone walks into a room. It’s all unconscious, and we can’t help it.

I also have severe combined-type ADHD. Some experiences are literally painful for me because of how taxing they are on my attention and cognitive load. I’m acutely sensitive to friction, clutter, noise, confusion. The things that bad design inflicts on everyone, I feel at a higher volume. I design because I’m sensitive to these patterns. What most designers theorize about, I feel.

And the ADHD gave me something else: years of building mental prostheses. If it’s out of sight, it’s literally out of mind for me, so I keep post-it notes on my desk for today’s tasks. “Out of sight, out of mind” becomes: what MUST we show vs. what we want to show vs. what we could show. Then cut everything but what we must show.

Design is prosthesis for human cognitive functions and limitations. I know that because I’ve been building prostheses for myself my whole life. The curb cut effect: what I built for my own cognitive accessibility turned out to help everyone.

I’ve come to think of this as a curb-cut effect. The cognitive constraints I designed around — limited working memory, sensitivity to disfluency, impatience with unnecessary friction — are constraints every user has. Mine are just louder. A design that works for the constrained case tends to work better for everyone.

The five PFD layers draw from two largely distinct bodies of cognitive science — perception psychology and ADHD cognitive neuroscience — connected here for the first time. The citation lists share no duplicate entries.

The Conductor Model

I don’t build funnels. I conduct.

Think about a musician who’s reached unconscious mastery. They stop thinking about how things are done and focus entirely on how the music makes you feel. That’s the goal. Prime someone with a first impression, then conduct their thoughts and emotions through the experience. Design a mental waterfall.

There’s a flow of thought and perception: momentum, weight, cadence, tone, speed, context, priming, expectations. Like singing down a scale in front of a crowd and pausing before the last note. Everyone knows what comes next. That recognition, that pattern completion, is satisfying. That’s what a great CTA feels like: the only possible resolution to the experience that preceded it.

The User’s Mental Waterfall

The unconscious sequence of questions every visitor processes:

1. What am I trying to accomplish here? → Intent
2. Is there a clear path to that? → Navigation
3. What’s the point of this site? → Value proposition
4. For whom is this? → Do I belong here?
5. What is it trying to tell me? → Messaging: ew or cool?
6. What is it trying to get me to do? → Call to action
7. How do I feel about doing it? → “Prove you’re worth my time” or “Tell me more!”
8. Where’s the win-win? → Mutual value
9. Does this give me what I want? → Delivery

The hero section is the activation point; the table visit at 1:10. If it answers questions 1 through 5 in the first second or two, the user is primed. Then you conduct them through 6 through 9 as they scroll. If the hero fails, it’s the lights slamming on.

My improv teacher gave me a formula: (Listen) × (Act + React). If Listen equals zero, if the user isn’t attending, then everything else multiplies by zero.

The Framework

Five layers as a dependency stack. Each depends on the layer below it. Fix them in order, because upstream failures degrade everything downstream. But think of it less as a stack and more as a composition. The foundation sets the key. Each layer adds another instrument. If the foundation is broken, the whole piece is noise.

In predictive processing terms (Clark, 2013; Hohwy, 2013): when a design matches the brain’s predictions at every layer, processing is fluid and unconscious. Autopilot. When a layer fails, it generates prediction error that demands conscious attention. Good design means each layer confirms predictions so the brain’s resources flow toward the decisions that matter, not toward deciphering the interface.

The Diagnostic

How I actually diagnose a site. Not a checklist; a process.

Step 1: Feel (Pre-Verbal)

Arrive at the page and let the emotional response fire before Wernicke’s area translates it into words. “This makes me feel X.” The feeling comes before the language. System 1 before System 2. My ADHD sensitivity is an advantage here; I feel cognitive load and friction more acutely than most users. The emotional response IS the diagnostic instrument.

Step 2: Unpack

“Why does this feel X?” Now I’m looking for the systematic rule that’s broken: don’t overload, be specific, get to your point, signal-to-noise ratio, everything has a purpose, this should feel X. Does it?

Step 3: Diagnose

Step 4: Prescribe

Fix the lowest failing layer first. Always. Fixing decision architecture when the first impression is broken is like optimizing a trail that nobody enters.

The 5-Minute Perception Audit

The Generative Protocol

The Diagnostic decodes failures. The Generative Protocol derives solutions. Evaluation and generation are different cognitive operations. Evaluating a proposal against PFD catches surface violations. Deriving a solution bottom-up from PFD’s constraints surfaces requirements that no existing proposal addresses.

1. State the Design Problem in terms of user experience, not missing features
2. Work Each Layer Bottom-Up. For each: state the constraint, state the violation, derive the MUST requirement, label R[n]
3. Accumulate Requirements. R1–R5, all non-negotiable. Lower layers win conflicts.
4. Derive the Solution. What satisfies R1 AND R2 AND R3 AND R4 AND R5?
5. Test Against Proposals. Which requirements does each proposal fail?
6. Identify the Gap. Requirements no proposal satisfies = where the non-obvious solution lives

The Gate

Run before shipping. Each layer is pass/fail. If any layer below L4 fails, do not ship.

The Communication Hierarchy

Music, humor, and stories: three modalities for changing perception, ranked by power.

Music (Most Powerful)

Music communicates emotion and bypasses rational processing entirely. Koelsch (2014) showed music modulates subcortical emotion structures rapidly, potentially prior to conscious evaluation. Juslin & Västfjäll (2008) identified six mechanisms for music-evoked emotion, the majority bypassing conscious processing. In design: typography rhythm, spacing cadence, visual flow, animation timing, scroll momentum. The “music” of a page controls pacing and emotional state.

Humor (Second)

Jokes change how people feel about a thing by pointing out absurdity and giving permission to think without rigidity. A joke is: setup creates expectation, punch reveals a pattern that either aligns with your self-image or doesn’t, and the anxiety between amplifies the payoff. Laughter is the anxiety being released. Three comedy escalation patterns map to UX: linear repetition (brand consistency), sequential escalation (funnel narrative), exponential escalation (hero → proof → testimonial → CTA).

Stories (Third)

Green & Brock (2000) demonstrated “narrative transportation,” where absorption into a story produces belief changes. Two meta-analyses confirm (van Laer et al., 2014; Braddock & Dillard, 2016). The story isn’t the events. It’s how someone changes from A to B. The user is the protagonist. “I have a problem” → “This solves my problem.” The hero is the inciting incident. The CTA is the climax.

Websites as Relationships

Users evaluate websites the same way they evaluate people. Same unconscious process, same red and green flags.

The trust-killer that operates before anyone reads a word: disrespect. A site that doesn’t treat users with respect, their time, their intelligence, their needs, triggers the same rejection as a person who acts like you’re lucky to interact with them.

The Responsibility

If we’re engineers of thoughts and emotions, we need a Hippocratic oath.

The distinction: The designer controls the perception layer: presentation, UX, visual hierarchy, emotional messaging, activation points. The organization controls the value layer: product quality, pricing integrity, fulfillment, service. PFD’s ethical mandate is to align perception with reality.

When the product is good and the perception is bad (the Simply Smart Home case) PFD closes the gap upward, bringing perception up to match genuine value. That’s not manipulation. That’s correction. The violation occurs in the opposite direction: inflating perception above reality.

The Three-Part Ethical Test

1. The Alignment Test: Does this design bring perception closer to reality, or further from it? (Berdichevsky & Neuenschwander, 1999)
2. The Sincerity Test: If the user fully understood what this design choice does, would they feel served or exploited? Consumers don’t object to persuasion. They object to persuasion perceived as insincere (Friestad & Wright, 1994; Campbell & Kirmani, 2000).
3. The Golden Rule: Would I, as the designer, consent to being influenced by this technique if I were the user?

Dark patterns (Brignull, 2010; Gray et al., 2018), nagging, sneaking, forced action, are the cheap laughs of design. They convert short-term and destroy trust. PFD is the expensive laugh: genuine emotional resonance that compounds.

Relationship to Value Sensitive Design

PFD’s ethical framework sits within the HCI tradition of Value Sensitive Design (Friedman, Kahn & Borning, 2002). Knobel & Bowker (2011) extended this to ask how values get embedded into designed systems, not just studied after the fact, but actively inscribed through design decisions that privilege certain stakeholders, workflows, and outcomes over others.

PFD’s perception layer is exactly such an inscription point: every visual hierarchy, every default, every activation point encodes a judgment about whose attention matters and what they should do with it. Where VSD asks broad questions about which values a system embeds, PFD’s ethical mandate is narrower and operational: does this specific design bring perception closer to reality, for this user, in this context?

What PFD Doesn’t See

Every framework has blind spots: things it makes visible and things it occludes. Knobel’s concept of ontic occlusion (how one representation of reality blocks another from being seen) applies to PFD itself. Naming these isn’t a weakness. It’s a quality criterion.

Structural power dynamics. PFD’s ethical test says “align perception with reality.” But it doesn’t interrogate who decides what “reality” is. When SSH’s product page undersold a good product, the “reality” was clear: the product was better than the perception. But what about a company whose product is mediocre and whose “reality” is genuinely ambiguous? The alignment test assumes a knowable ground truth. In practice, “reality” is contested, and the designer choosing which reality to align with is itself a power move that PFD doesn’t surface.

Individual vs. collective. PFD models one user’s cognitive journey through the mental waterfall. It doesn’t model what happens when thousands of users are conducted through the same waterfall simultaneously. Individual perception optimization can produce collective harms: engagement loops, attention extraction at scale, homogenization of choice. PFD’s unit of analysis is the user. The effects that emerge at population scale are outside its frame.

The fluency trap. Processing fluency is L2’s core mechanism: “if it’s easy to process, it feels true.” That mechanism is value-neutral. It works for honest brands and dishonest ones. A fluent lie is more persuasive than a disfluent truth. PFD’s ethical tests catch deliberate manipulation, but they don’t address the structural problem: by optimizing for fluency, PFD makes it harder for users to engage the critical evaluation that would catch deception. The very thing that makes PFD effective (reducing prediction error, keeping users in autopilot through the desired path) is the same thing that can prevent them from stopping to question. This is ontic occlusion baked into the mechanism.

The designer’s own perception biases. PFD’s diagnostic starts with “Feel”: the designer’s pre-verbal emotional response as the primary instrument. But that instrument is calibrated by the designer’s own culture, class, neurotype, and aesthetic history. I name my neurodivergence as an advantage (heightened sensitivity to friction). But heightened sensitivity to friction could also mean under-sensitivity to other things: social warmth cues, cultural signifiers outside my experience, accessibility needs I don’t personally feel. The diagnostic instrument has its own biases, and PFD doesn’t yet have a protocol for calibrating them.

The Results

Simply Smart Home

The problem: $1.7M annual revenue. Template site. No unified branding. Feature-focused marketing. $150+ price point with a presentation that said “$15 knockoff.”

Every layer was failing: site didn’t answer basic questions (Foundation), no faces or warmth (L1), template at premium price (L2), feature specs instead of emotional connection (L3), homepage structured around company, not user goals (L4).

What I did: Unified brand system. Emotional messaging: “stay connected, even when you’re apart.” Hero sections with faces. Consistent collateral across every touchpoint. 4-year systematic brand infrastructure build.

• 233% year-over-year sales increase
• $1.7M to $5M annual revenue
• Disney licensing, Costco pallets, Walmart shelf space
• Competitor stole the tagline two years later

iO Theater

“Doesn’t look like a real theater.” Full redesign: dramatic photography, clear event navigation, integrated payment. Online ticket sales 50% → 75%. Same shows. Different perception.

VacuumSealersUnlimited

Embedded design partnership, 5+ years. $5M+ annual revenue running on infrastructure I built and maintain. The long game. Perception compounds.

What’s Teachable

3 Hours with Business Owners

Hour 1: Business Design. Your business IS a design. Feedback loops between your vertical, users, and staff. I’d give them the tools to see their business as a system.

Hour 2: People as Research. Your staff and clients are your greatest resources. Design problem-solving is really just a lens of empathy. “For whom is this?” gives you directionality that frameworks can’t.

Hour 3: Tactics. The 5-Minute Perception Audit. The dependency stack. The red flags. Application, communication, process.

The Teachability Line

The lens (empathy, for-whom, design-as-problem-solving) is fully teachable. The pattern library (hundreds of projects of instant recognition) takes years. The sensitivity (neurodivergent tuning fork for bad design) may not be teachable. But the diagnostic process (Feel → Unpack → Diagnose → Prescribe) can be taught as a substitute.

The Improv Rules

My design practice runs on principles I learned from improv classes. These aren’t metaphors. They’re operational tools.

Open Questions

These are genuinely unresolved. A living methodology names its edges.

Cultural generalizability. The evidence base is predominantly WEIRD (Western, Educated, Industrialized, Rich, Democratic). The underlying cognitive architecture (predictive processing, working memory limits, first-impression formation) is expected to generalize. But the parameters (which visual patterns signal trust, what counts as “too much” cognitive load, how fluency maps to credibility) are culturally shaped. PFD has been applied in Western e-commerce and entertainment. How the 5-layer dependency order holds in collectivist markets, non-Latin typographic traditions, or markets with fundamentally different trust signals is an open question.

Measurement. PFD lacks a validated measurement instrument. The 5-Minute Perception Audit is a practitioner heuristic: useful, fast, and directionally correct, but without inter-rater reliability data. A diagnostic tool where two PFD practitioners independently audit the same site and converge on the same layer diagnosis would strengthen the methodology from “expert intuition with a framework” to “teachable, repeatable diagnostic.” This is the gap between a clinical art and a clinical science.

Temporal dynamics. PFD describes the first encounter well: activation, first impression, mental waterfall. But user relationships with products evolve. At visit 100, the mental waterfall is different. Prediction errors that activated on first visit are now expected. Fluency that felt trustworthy can become invisible (users stop noticing). Does the 5-layer stack apply the same way to retention and long-term engagement as it does to acquisition? The current framework is weighted toward first contact.

Accessibility as justice vs. fluency. PFD frames accessibility through Foundation (cognitive load reduction) and L2 (processing fluency). This produces good accessible design in practice. But disability justice scholars argue accessibility is a rights issue, not an optimization target. Framing it through “fluency” risks treating accessibility as something you optimize when the ROI justifies it, rather than something you guarantee because people deserve it. The framework’s utilitarian lens may not fully serve this.

AI-generated design. As AI tools generate more design artifacts, PFD’s emphasis on embodied sensitivity (the “neurodivergent tuning fork,” the pre-verbal Feel step) becomes harder to operationalize at scale. Can PFD’s diagnostic be automated? Can an AI run the Feel step? If not, the methodology has a ceiling on how far it scales beyond the individual practitioner. If yes, what’s lost when the diagnostic instrument is no longer a human body?

Appendix A: The Science

78 citations across cognitive psychology, HCI, behavioral economics, design theory, STS, music cognition, narrative psychology, and ADHD cognitive neuroscience.

Appendix B: Glossary