The deliberate design practice blog
Essays and playbooks on deliberate design practice: how to run design challenges, write UX case studies, and apply persuasion psychology to real product work.
Onboarding flow: the first choice decides whether the user comes back
An onboarding flow is the first part of a subscription, not a toll before it. On a coffee subscription that loses 35% of new subscribers after the first box: ask for answers the person gets back, put the heavier investment after the reveal, and leave the next choice open.
· Flamel
The HEART framework on a real brief: pick one signal per letter, drop the rest
The HEART framework is Google's five categories of UX metrics: Happiness, Engagement, Adoption, Retention and Task success. Filled in whole, it decides nothing. On a film club's membership page: one signal per letter, the reason the others are out, and the letter that decides.
· Flamel
AI UX designer: what the job asks for beyond prompting
AI UX designer names two jobs: designing a product with a model inside, and designing with a model at your side. Both ask for judgement about trust, error and control. How to read the listings and show you have it.
· Flamel
UX design projects with a real brief: what to practise when AI does the screens
A practice project trains judgement only if it brings a brand with constraints, a specific user and a number to move. What each kind of project gives you, when to skip it, and four briefs that carry all three.
· Flamel
Habit-forming products: design the return, not the compulsion
Compulsion makes leaving expensive. A healthy habit makes coming back worth it. The difference is decided on one screen: what your product shows someone who returns after two weeks away.
· Flamel
Product page design: when the decision is the size, not the photo
On a page that sells something people cannot try on, the photo rarely stops the sale. On a €400 ring it is the size: answer it on the page, and say what happens if it is wrong.
· Flamel
UX strategy for a product designer: deciding which moment to fix first
UX strategy is the order in which you fix the customer's moments, and the argument that holds it. Three questions set that order, and on a newsroom's paywall they put the first week before the wall.
· Flamel
Designing for AI: the hard problems aren't in the chat box
AI UX is behaviour design: define autonomy, preview, provenance, undo, and failure—not a chat box. Set contracts to match risk and moment, keep policy and thresholds human-owned, and prove them with time-to-first-correct and reversal-within-30s.
· Flamel
Laws of UX are a list. Behaviour design is knowing which one is in play
The laws of UX are easy to quote and hard to apply. The skill is knowing which one governs the screen in front of you: on a gift finder for a buyer in a hurry, Hick's law decides how many questions to ask, not how many buttons to show.
· Flamel
UX design practice with an AI draft to beat
UX design practice in 2026 starts where the model stops: a correct, generic draft and sixty minutes to beat it with one decision you can defend. How to set the exercise up on your own, and how to tell whether you won.
· Flamel
Design engineer: the role AI is creating, and whether a product designer should move into it
A design engineer designs in code, so the interface they hand over is the one users get. AI made building cheap; what the role asks, what a product designer keeps, and what isn't worth learning.
· Flamel
Behavioral design on an abandoned cart: the trigger is already internal
Behavioral design is choosing the one trigger that moves someone and designing for that. In an abandoned cart the trigger is already internal, a doubt: name it, then match one explicit prompt to the buyer's state.
· Flamel
Checkout UX for a limited drop: a simplicity problem, not a layout problem
In a limited drop, the pay step has only ability left: the total, the time to pay and the taps on a phone decide who keeps a piece. How to cut each one, and what not to add.
· Flamel
Growth design course or daily practice: what each one trains
A growth design course gives you the frame; daily practice gives you the repetitions. What Reforge, Growth.Design, a bootcamp and daily practice each train, and when you need none of them.
· Flamel
UX case study template: write it from the brief, not from the screens
A UX case study template that starts from the screens produces case studies that look alike. Start from the brief: the brand, the person, the number and what you left out. Nine fields, and the template filled in on a real practice brief.
· Flamel
AI prototyping: what to hand to the model and what to decide yourself
AI prototyping hands you a correct screen in minutes. What it does not decide is who the screen is for, which moment it fixes and what it leaves out. The split, the rule behind it, and the AI draft in a real practice brief taken apart.
· Flamel
UX metrics a designer can defend: one number per moment
The UX metrics a designer can defend are one number per customer moment and the sentence that ties it to the screen you changed. The five moments, the sentence, and a subscription quiz measured by who stays rather than who finishes.
· Flamel
UX designer, product designer or growth designer: what each one is paid to decide
UX designers answer for being understood, product designers for what gets built and why, growth designers for the number after launch. A map of the three roles, the same feature seen by each, and how to move between them.
· Flamel
The Hooked model, drawn on a real brief
Most Hooked loops fail because they skip investment; use Fogg's gate, state-aware actions, and tribe-fit variable rewards so each action stores value and strengthens the next trigger. Drawn on an indie game studio that builds in public, with wishlists and a community around its devlogs.
· Flamel
Multi-step forms: order the asks by commitment, not by fields
Order a multi-step form by commitment, not by the database: identity and intent first, what the choice needs next, payment last. Then read where people leave, especially at step two, and defend the order in review.
· Flamel
Countdown timer UX without fake urgency: designing a real drop
Use countdowns only when they enforce a real, verifiable limit—server time, live stock, or a cart hold tied to inventory. Anything else is a dark pattern; design urgency that is specific, checkable, placed at the decision, and measured with lift and regret.
· Flamel
Growth design: what a growth designer does that a UI designer doesn't
Growth design means answering for the number that moves after launch: you pick the moment that blocks growth, choose the lever that unblocks it, and measure lift and regret together. Here is what that looks like on a real brief, and how to practise it.
· Flamel
Teardown: Airbnb's social proof is doing three jobs at once
Airbnb's reviews module converts by bundling checkable trust, load-reducing pre-sorting, and timeboxed, listing-local demand cues at the booking moment. Copying stars without sourced claims, near-CTA placement, and a visible undo misses causal fit and lifts cancellations.
· Flamel
Psychology in UX design is the part AI can't copy yet
AI can draw any pattern; the hard part is choosing a premise and mechanism that fit this person, this moment. Design for causal truth: state checkable claims, place the trigger at ability, keep a visible undo, and measure lift with regret so clarity beats pressure.
· Flamel
Dark patterns vs persuasive design: where the line actually is
The line between dark patterns and persuasive design is not the technique. It is whether the claim moving the user is true, and checkable, at the moment of choice. How to test a nudge, measure it against regret, and defend the call.
· Flamel
The four kinds of friction, and which one you are actually removing
Simplifying UX works only when you name the exact friction you're cutting: mental, physical, time, or cost. Identify the blocking mechanism, state a testable hypothesis, and choose patterns that reduce that specific cost, not just steps or text.
· Flamel
Five things to never let AI decide
Never outsource five product calls to AI: choose the metric, draw the ethical line, decide exclusions, set brand voice, or promise outcomes. You own these upfront and instrument them, or you’ll pay in misaligned incentives, quiet drift, and broken trust months later.
· Flamel
How to critique an AI-generated screen
A generated screen is correct in everything and decisive in nothing. Critique it by naming the bet it refused to make: the hierarchy that ranks a goal, the objection answered at the commit point, the friction removed on purpose — and the smallest proof that would change your mind.
· Flamel
AI design tools in 2026, compared honestly
Compare AI design tools by where they enter your week and the testable asset they return, not by the demo screen. The ones worth money force a decision under a constraint and hand back evidence you can judge tomorrow; the rest decorate your indecision in higher fidelity.
· Flamel
The design brief template: every field, and why it's there
A design brief template is only useful if every field closes a decision. The eight fields that do it, what breaks when each one is missing, a filled-in example, and why the same template is also your best AI prompt.
· Flamel
Daily UI is not practice. Here is a 30-day plan that is.
Daily UI rewards pixel variety, not the judgment product work now demands. Train a daily loop that frames a problem, forces a trade-off, and defines a measurable check; AI can draft screens, but your advantage is a defensible decision.
· Flamel
How to prepare a design take-home assignment in the AI era
Make your design take-home explain the decision behind the screens. Show what you chose, what you rejected, and what you would check before shipping.
· Flamel
What product design interviews actually test in 2026
In 2026 design interviews, surfaces are free and only priced judgment counts: a mechanism at a gate, a bound, a metric with bite, a read window, and a stop-loss with a revert. Panels test if you can turn ambiguity into a priced, local, reversible bet they can trust.
· Flamel
Why unsolicited redesigns teach you nothing (and now prove even less)
Unsolicited redesigns prove little now that AI makes pretty screens cheap; what counts is a priced decision under constraint tied to a behavior. Replace galleries with reversible bets: gate, metric, MDC, stop‑loss, window, and receipts.
· Flamel
Six case study red flags reviewers notice in the first minute
Reviewers buy decisions under constraint, not decorated timelines. Make your opener a priced, reversible bet with a bound, mechanism tied to the moment, success metric, receipts, MDC, read window, and a revert path so your case reads as accountable authorship over ornament, even under NDA.
· Flamel
How to build a UX portfolio when your best work is under NDA
An NDA blocks brand assets, not your ability to prove judgment: publish the decision, the constraint, the lever, the success check, and the revert without logos or raw data. Ship a one‑page case or a practice project with redacted surfaces and indexed deltas to show priced, reversible bets.
· Flamel
When everyone's portfolio looks good, how do you prove the judgement was yours?
Polish is now cheap; judgment under constraint is the scarce signal. Make your portfolio a traceable ledger of decisions—choice set, lever, bounds, checks, and reverts—so reviewers can price risk and trust your authorship even when surfaces look the same.
· Flamel
How to present design work to executives
Executives don’t buy screens; they buy reduced risk, faster impact, and controlled cost. Lead with the decision, mechanism, evidence, and stop-loss—name the moment, metric, window, and revert—then show surfaces; AI made visuals cheap, so credible, bounded bets are your edge.
· Flamel
What to do when the PM overrules your design
Designs lose to PMs not on pixels but because claims use mismatched units; convert vibes to comparable, time-bound bets with mechanism, moment, metric, and a revert. Buy the smallest check, price risk and effort, and win or learn fast—even against AI variants.
· Flamel
Turning taste into an argument: the vocabulary designers are missing
Design decisions should state a mechanism and a check, not a mood. Name the perceptual, memory, or motivational cause; turn it into a falsifiable, cross-functional criterion tied to a moment and metric, then measure the smallest behavior change while AI handles execution.
· Flamel
How to defend a design decision
You do not lose design reviews because your work is weak. You lose them because your argument is about the screen and theirs is about the business. And now that anyone in the room can generate the screen, the argument is not a complement to the job. It is the job.
· Flamel
How to write a design rationale (and the one sentence that does the work)
Design rationale is one falsifiable line: change X because Y, and we will know by Z. Write it first, tie Y to user behavior and Z to a near-term decision, cut artifacts and vibes, and record only the bet and outcome so teams decide and learn fast.
· Flamel
Teardown: what Duolingo's streak actually does to you
The streak is not a reward. It is an asset the product lets you build and then holds. Here is the mechanism underneath it, why copying the number never works, and the exact point where it stops being fair.
· Flamel
The Fogg Behavior Model, applied to real product decisions
A behavior that is not happening is missing motivation, ability, or a prompt, and the three failures need three different fixes that cannot be substituted for each other. The Fogg Behavior Model is worth the hour because it forces a team to name which one is missing before anyone starts building the answer.
· Flamel
Brief an AI the way you'd brief a junior designer
The prompt that works isn't longer than the one that failed, it's shaped differently: it has the shape of a brief. Five fields decide the work, and if you leave one blank the model fills it silently and hands you a finished artefact instead of a question.
· Flamel
The product designer's AI workflow: what to delegate and what never to
Most AI workflows for designers are tool lists, and tool lists expire. This is the other kind: a division of responsibility that says exactly what the machine produces, what you decide, and which handover is the one that breaks everything when you get it wrong.
· Flamel
The new seniority ladder: what mid and senior mean now
The ladder everyone climbed measured volume and finish, and it stopped working the month competent output got cheap. The one underneath it measures how far your decisions reach — and it explains what mid and senior actually mean now.
· Flamel
What AI got good at, and what it still can't do
Most inventories of what AI cannot do in design are written to reassure, and they age badly. This one concedes the whole left column first, then draws the line where it actually falls: not at difficulty, but at the decisions somebody has to answer for.
· Flamel
How to compete with AI as a designer (you don't)
You cannot win a race to produce the artefact, and the prize is the input whose price is collapsing anyway. The position that is open is the one that says which of five directions ships, names the mechanism behind it, and carries the number that would prove it wrong.
· Flamel
The three design skills that appreciate as AI gets better
Future-proofing is not about finding the corner AI cannot reach. Three skills get more valuable every time the models improve: naming the mechanism behind a decision, defending it out loud, and directing the machine instead of racing it.
· Flamel
Will AI replace product designers? What it already replaced
AI has already automated first drafts, filler copy, variants, sample data and admin in product design, but not the judgement of which problem to solve, how to sequence, what to measure or what to cut; that is now the job.
· Flamel
Is it still worth being a product designer in 2026?
The honest answer is yes, but not to the same job. Here is exactly which part of design work disappeared, which part got scarcer, and the three skills that are worth more now than they were three years ago.
· Flamel
UX case study examples: what the strong ones actually show
Most UX case study examples teach the wrong lesson. Here is what separates a case study that gets you hired from a well-designed slideshow, with the structure to write one.
· Flamel
How to run a 60-minute design challenge (and why the timebox is the point)
A practical protocol for daily UX design challenges: how to split the hour, what to cut when you run out of time, and the four failure modes that make most design practice useless.
· Flamel
Where to find realistic design briefs to practice with
Design brief generators mostly produce prompts, not briefs. Here is what a brief needs to contain to be worth your time, where to get real ones, and a template for writing your own.
· Flamel
Figma practice projects that are actually worth your time
Most Figma practice produces tool fluency and nothing else. Six project types that build judgement instead, and the reason component-level exercises stop paying off so quickly.
· Flamel
The persuasion principles every product designer should know
Cialdini, Fogg, Eyal and the Heath brothers in one map: what each framework is actually for, which one to reach for when, and where each crosses into dark patterns.
· Flamel