Lab note 00 A field guide to the steps between wanting and doing

The case for good friction

Product design has spent two decades sanding away every obstacle. Most of that was a gift. But a few obstacles were never obstacles — they were the moment where you decided, consented, noticed, or got a second chance. And some of the friction that remains isn't protecting anyone at all; it's quietly keeping people out.

Slide toward zero and the words lose traction. Slide too far and they lock up.

Open the friction bench
A curve: harm falls as friction rises, then exclusion rises. The good zone sits between. good friction unguarded harm exclusion &fatigue too easy too hard
Interpretation A sketch, not data: two costs pulling in opposite directions. Where the zone sits depends on what can be undone.

01 · The material

Friction is a design material. It comes in three grades.

“Friction” is used as an insult in product meetings: anything that slows a person down. That framing hides a question that matters more than speed — whose interest does the slowdown serve, and can everyone get through it? Ask that, and the same-looking step splits into three very different things.

Protective

Slows you down for you, roughly in proportion to what you could lose. An undo window, a review screen before money moves, a pause before a message reaches four hundred people.

Test: would you thank it on your worst day?

Manipulative

Slows you down for them — and usually only in one direction. Joining takes a tap; leaving takes a phone call. Accepting is a big button; refusing is five screens.

Test: is the effort symmetrical?

Excluding

May be well-intended, but treats one kind of body, memory, device or schedule as the default. A puzzle with no alternative; a booking that requires a smartphone; office-hours-only.

Test: is there a second way through?

02 · The friction bench

Pick a situation. Turn the dial. Watch what breaks.

Ten everyday design situations, each built at five friction settings. The four gauges are a Design scenario model — reasoned estimates to make trade-offs visible, not measurements. Where a real study or rule exists, it's cited under the card.

·

    Use the slider, arrow keys, or the step buttons.

    The case for less

    The case for more

    03 · Six principles

    How to put friction where it belongs

    1. Scale friction to irreversibility, not to importance.

      A decision can matter enormously and still deserve no friction if it's easy to reverse. Moving money, deleting data and making legal commitments are different: the web's accessibility guidelines ask that such submissions be reversible, checked for errors, or reviewable before they're final. WCAG 2.2 · SC 3.3.4

    2. Prefer a window to a wall.

      An undo period after the action keeps nearly all the speed and recovers most mistakes. A dialog before the action interrupts everyone to catch a few. Where reversal is possible, build reversal first.

    3. Be rare, so you're read.

      Confirmation dialogs work when they're unusual. Nielsen Norman Group warns against using them for routine actions: people learn to click through, and the dialog stops preventing anything. NN/g on confirmation dialogs

    4. Name the consequence.

      “Are you sure?” is a lazy question. “This permanently deletes 214 photos shared with 3 people” is friction that does work: it gives the pause something to think about.

    5. Make the exit as easy as the entrance.

      Asymmetry is the fingerprint of manipulation. If joining is one tap and leaving is a phone call, the friction isn't protecting the person — it's protecting revenue from the person.

    6. Never let a safeguard become a gate.

      A security step that requires remembering, transcribing or solving something must offer another route. WCAG's authentication criterion specifically names allowing password managers and paste as acceptable help. WCAG 2.2 · SC 3.3.8

    04 · Telling them apart

    Five questions that separate a guardrail from a trap

    QuestionProtectiveManipulativeExcluding
    Who benefits from the pause?The person paying itThe businessOften nobody; it's an assumption
    Is it symmetrical?Same effort either way, or more on the riskier sideEasy in, hard outHard for some people in both directions
    Is it honest about why?States the consequence plainlyUses guilt, false urgency or hidden defaultsUsually silent about who it fails
    Is it proportionate?Grows with irreversibilityGrows with how much revenue is at stakeFixed, regardless of stakes
    Is there another way through?Yes — undo, help, alternativesDeliberately obscuredNo — that's the defect

    05 · Field test

    Sort the specimens

    Nine real-world patterns. File each one as protective, manipulative or excluding. Some are deliberately close calls — the explanation matters more than the score.

      06 · Reversible workshop

      Build a better interface. Then undo your way to the best one.

      Start from the frictionless version of a situation and add design moves — including a few dark ones, so you can see what they do to the gauges. Every change is in an undo history. When you're happy, export a decision card to share with a team.

      Design moves

      Protective

      Manipulative

      Excluding

      07 · Against easy

      Four critiques of convenience culture

      Interpretation These are arguments, not findings. Each comes with the strongest objection I know.

      i.

      Convenience is measured at the wrong moment.

      Products tend to count ease at the instant of action — the tap, the checkout. The cost of an unconsidered action arrives later: the return, the regret, the subscription found three months on. A metric that only sees the first moment will always vote for less friction.

      But: people are entitled to make quick choices they'll sometimes regret. Designing for “future you” can slide into paternalism if it isn't paired with easy reversal.

      ii.

      When the step disappears, so can the decision.

      Consent isn't a checkbox; it's a moment of noticing what you're agreeing to. Remove every pause and you can end up with actions nobody exactly chose — tracking switched on by default, purchases from a pocket, posts shared on reflex.

      But: pauses that appear everywhere become noise. Consent theatre — a banner on every page — may produce less real choice than one good default.

      iii.

      Effort isn't removed. It's moved.

      A frictionless front end often relocates the work: to support staff untangling mistakes, to the person reconciling accidental charges, to whoever must explain to a relative why the money went to the wrong account. Ease for one party can be a bill for another.

      But: the reverse is equally true. Every approval layer and verification step moves effort onto the people least able to absorb it.

      iv.

      Some effort is the point.

      Booking a slot at a museum, writing a letter rather than a reaction, walking to a place — sometimes the effort shapes the attention we bring. Flattening every experience into the same instant tap can quietly flatten what the experience was for.

      But: “effort is meaningful” is easy to say when the effort is optional for you. For many disabled people, convenience isn't laziness; it's access that was fought for. Romanticising friction is its own kind of exclusion.

      08 · Sources & labels

      What's evidence here, and what isn't

      • Evidence A claim backed by the linked source.
      • Interpretation An argument or framing. Disagree freely.
      • Design scenario Invented situations and gauge values used as a thinking tool. Not measurements.
      1. W3C WAI — Understanding SC 3.3.4: Error Prevention (Legal, Financial, Data)Submissions with legal, financial or data-loss consequences should be reversible, checked, or confirmable before they're final.
      2. W3C WAI — Understanding SC 3.3.8: Accessible Authentication (Minimum)Logins shouldn't require remembering, transcribing or puzzle-solving without an alternative or assistance such as paste and password managers.
      3. Nielsen Norman Group — Confirmation Dialogs Can Prevent User Errors (If Not Overused)Confirmation dialogs lose their power when used for routine actions; they should be specific.
      4. Nielsen Norman Group — 10 Usability HeuristicsError prevention and user control (undo) as long-standing usability principles.
      5. TechCrunch (24 Sep 2020) — Twitter's “read before you retweet” promptTwitter reported that people opened articles 40% more often after seeing the prompt. Company-reported figures, not an independent study.
      6. US FTC — Bringing Dark Patterns to Light (staff report, Sept 2022)Catalogue of manipulative design tactics.
      7. Alston & Bird — FTC Click-to-Cancel Rule Vacated (July 2025)The Eighth Circuit vacated the FTC's amended Negative Option (“click-to-cancel”) Rule on 8 July 2025, on procedural grounds, before it took full effect.
      8. Deloitte India — Tax alert on the Guidelines for Prevention and Regulation of Dark Patterns, 2023 (PDF)India's CCPA notified the guidelines on 30 November 2023, identifying 13 specified dark patterns.
      9. Deceptive Design (Harry Brignull)Long-running catalogue of deceptive design patterns.

      Scenario gauges, verdicts and workshop scores are a deliberately simple model. Real products should be tested with the actual people who use them — especially people using assistive technology.

      Last step, deliberately

      Have a thought, challenge, or idea? Write to Sahil.

      Found a scenario where the bench gets it wrong? Know a protective step that deserves a place here — or a “safeguard” that locked someone out? This is the one bit of friction on the page worth adding.

      Opens your own email app. Nothing is sent or stored by this page.