A climber gets stuck on a grade and has no idea why. SEND IT films the attempt, shows them what their body did the moment their fingers opened, and tells them what to try next. I designed it as a full system – until I realised nobody opens a desktop while standing at the wall.
A climber stuck on a grade knows exactly one thing: that it did not work. They did not see themselves. They do not know whether the hip drifted, whether the foot was too high, or whether there simply was not enough left in the forearm. So they try again, the same way, and fall the same way. With nobody watching from the outside, the fifth attempt is identical to the first.
I built a system that knows everything about your climbing.The moment the project changed direction
Nobody opens it while sitting on the mat.
The first version of SEND IT was a desktop product called Performance Command Center – eight sections in a left rail, a nine-card grid on the home screen, HRV, SpO₂, lactic acid, a smart coach, a technique radar. It answered every question a climber might have. Its problem was not the answers. It was the room.
I built a system that answers every question a climber could have. I never asked when they ask them. The answer is: standing at the wall with chalk on their hands – and that is exactly where a desktop does not exist.
Eight rail items and nine home cards mean nothing was prioritised. When everything matters equally, the climber is the one who has to decide – and that is my job, not theirs.
| From the system | Verdict | Why |
|---|---|---|
| Readiness score | Kept – promoted | It is the first thing a climber wants to know walking in, so it moved directly under the header on Home. |
| Muscle load map | Kept – given a job | On the dashboard it was looked at. In the app the loaded muscles choose the recovery exercises. |
| Beta, step by step | Kept – redrawn | The ten-step rail became annotation drawn on the climber's body, on the frame where the move failed. |
| Avatars, badges, tokens | Kept – relocated | Gamification got its own screen instead of sitting next to the username competing with a profile photo. |
| The route page with hold detail | Kept – became the centre | On desktop it was a column beside the data. In the app it became the surface you work on. |
| HRV · SpO₂ · heart rate · energy | Demoted | Moved from a headline dashboard card to a second layer inside the Body tab, behind a watch. They explain readiness; they do not replace it. |
| Lactic acid | Cut | The one reading with no equivalent on a watch a climber actually wears. |
| Calorie intake · macros · body weight | Never added | Calories burned stayed – they measure what you did. Weight measures what you are, and in a strength-to-weight sport that difference matters. |
| The eight-item left rail | Cut | Five tabs. Anything that did not deserve a tab went into a menu, not into a ninth rail item. |
| The nine-card home grid | Cut | Nine equal cards is nine unranked questions. Home became an ordered answer to one: what should I do right now? |
| “Plan Next Session” as a global CTA | Cut | A button that plans a session you have not had yet. Replaced by Scan a problem – the only thing you can actually do standing in a gym. |
| Setter tools · gym signage · marketplace seller side | Out of scope, stated openly | Real surfaces this product would need. Naming them protects the work in review far better than quietly omitting them. |
Choosing the primary persona settled almost every argument that followed. A product for the climber who is improving can celebrate sends. A product for the climber who is stuck on a grade has to make failure legible, and has to stop treating a rest day as a lapse.
Climbs V4, two or three sessions a week, eighteen months in. Owns a phone and no tripod. Has fallen on the same move six times today and cannot tell you why – not because they lack effort, but because nobody in the gym can see their own hips. They do not want to be told they failed. They want to be told what moved.
Primary User
“How the hell do I send this thing?”
When I fall on the same move six times, I want to see what my body actually did, so I can change one thing instead of just trying harder.
When I arrive at the gym, I want to know whether today is a session or a rest day, before I have warmed up and committed.
When I finish, I want one number that moved, and I want to trust that it moved for a reason I agree with.
The lo-fi set was not a sketch of the hi-fi set. It was where the product got three of its hardest ideas wrong in public, cheaply, and where they got fixed.
Seven of the nine findings were not about how a screen looked, but about what it did. That is why the answer was a different platform rather than a restyle. The structural seven:
Lo-fi, rounds 01–06 – home · the hands-free standby screen · the phone-refereed challenge. The last two did not survive.
Lo-fi is where you get to be wrong for free. All three of these survived six audit rounds and were killed on the way to hi-fi – each one because it promised something the product could not actually deliver.
N
Block Gym
N
Tracking you
Hi-fi, round 16 – the same three, resolved. Standby became a shutter you press before you climb. The referee became an honour system that says so out loud.
In lo-fi every screen was boxes and hierarchy. This stage settled where the photograph sits, where the explanation sits, and where colour is allowed to appear at all.
Each of these was a live disagreement during the work, not a rationale written afterwards. Each one names what was killed, because a decision you cannot name the loser of is not a decision. The colour of every card matches the colour of the screen it lives on, in the strip below. A dozen smaller calls were made the same way, and each one is annotated on the screen it belongs to.
The three collapsed into one screen with a popover. Then the popover turned out to be rendering over a different screen from the one the button lived on – which is simply wrong. Screen 08 is now the same screen as 07, exactly; the only difference between them is the popover sitting on top.
A popover is a layer over the context it came from, not a navigation destination.
Layer, not destinationA fall raises the analysis popover; a send raises the SENT popover with a Lottie. Exactly the same pattern for both branches. Breaking the palette for the celebration was given up in exchange for a consistent rule, and the animation carries the drama instead of the colour.
Consistency is worth more than one spectacular moment.
Consistency over spectacleIt was a promise the product cannot keep. A bad measurement makes the climber argue with the score instead of working on their feet. Now the climber reports: nobody is listening – mark it honestly or it measures nothing. It forced a second decision: a self-reported challenge cannot feed leaderboards, so it stays personal, and the boards count only what the gym can verify. Without that, the honour system is a competitive exploit.
An honest system with a stated hole beats a dishonest one that looks sealed.
Honesty over automation
Drag · scroll · arrow keys – FIG.03 the six screens, colour-matched to their card
Screen 14 says it plainly: readiness fell 67 → 38, and all 34 attempts moved it, not only the 3 that worked. This is the line that separates the rating, which measures outcome, from the body, which measures load. The same position drives challenges: “60 attempts this week” makes failure pay directly.
Physiology does not care whether you topped out.
Load, not outcomeAn app that contradicts itself teaches people to ignore it. The streak counts sessions, not days – “third session running”, with copy saying explicitly that rest days do not break it. The deload week still earns XP and still advances challenges, so backing off never quietly costs you anything.
Never punish the thing you just prescribed.
No contradictionCheck-in asks when you ate, not how much. No calorie counting, no macros and no body weight – anywhere in the product. Calories burned stay, because they measure what you did, not what you are. Climbing is a strength-to-weight sport with a documented eating-disorder problem, so this is a safety decision rather than a simplification. The recovery plan says “within two hours” – no number, nothing to count.
Some questions are cheap to ask and expensive to answer.
A safety decisionThe second version of the model deducted points for a failed attempt. It was tidy, and it was wrong: it penalised projecting – the single behaviour the product exists to encourage. The rating now moves on sends only and never falls. What a failed attempt does move sits in the other two numbers, and every screen explains the rules of the number it shows.
| Number | What it means | What moves it |
|---|---|---|
| Climbing rating 1408 → 1436 · V4 | How well you climb. The only number that is about climbing. | Sends only. It never falls because an attempt failed – a fall is training, not a loss. |
| XP 2,848 | App currency. Says nothing about how well you climb, and it is a balance, not a level, so it never reads as a ladder. | Every attempt, every challenge, every published beta. |
| Readiness ~38 | Fatigue, not worth. The tilde stays even with a watch on – readiness is computed from measurements, it is not itself measured. | Every attempt, sent or not. All 34 moved it, not just the 3 that worked. |
↑ Problems carry ratings too, so the problem is the opponent and gym grades calibrate themselves against real send data instead of a setter's opinion.
Everything below is the real interface, rendered in the page at 390 × 844 and scaled down. Each one says what it is for before it says why it looks that way – because that is the order the decisions were made in.
Home to explanation. The climber presses the shutter standing next to the phone, climbs, comes back to one question – and then to the move that failed, drawn on the body.
N
Block Gym
See readiness, today's goal and your last session – then tap Scan a problem.
The order follows what a climber wants to know walking in: readiness first, because it decides how the session goes. Then the goal, then the clip from last session – you see yourself climbing, which is the emotional hook of the screen. The header is three layers: the app (menu, logo, photo), the place (avatar + gym name, centred together), and the state (rating, grade, XP). One primary action.
Analyse raises the result in place, without leaving the screen you were on.
This is the same screen as 07, exactly; the only difference is the popover. The result opens where you tapped. It gives the bottom line – the move that failed, the XP, and “Climbing rating unchanged”, because a fall is training, not a loss.
Step through the sequence, explained on the body, on the move that failed.
The explanation is drawn on the frame, not captioned beneath it: a teal skeleton tracing her limbs, a dashed cue toward the next hold, labels anchored to the joint they describe. That is how a person explains climbing – by pointing at the body. Same climb as the card on Home; one clip, two depths.
Point the camera at the QR on the wall and the problem opens.
The consent sheet appears here, once per gym: faces blurred, analysis on-device, publishing asked separately – three lines and a button. A privacy gate shown after the camera already ran is not a gate.
DB
Explore the wall on the photograph – pan, zoom, tap a hold – then Record an attempt.
The photograph is the screen, not a thumbnail beside it. Explicit + / − controls exist because chalked hands miss pinch targets. Community data – send rate, who sent it, published beta – lives inside the problem, because this is where the decision gets made.
Tap a hold in the viewer and see it up close: name, type, size, orientation.
The close-up is a real crop of the same wall – the long red pinch in the middle of the line. A climber who reads “good for matches” before leaving the ground has already built half a sequence in their head. That is the whole job of this screen.
Set the phone down so the whole problem is in frame, then hit the shutter.
A full-bleed viewfinder, exactly like opening the phone's own camera. One centred instruction instead of an explainer card. The camera-flip control sits beside the shutter: front camera to watch yourself place it, rear camera to face the wall. A failing check still says what to fix.
N
Tracking youThe phone is down and rolling. Stop it on the way back.
The same viewfinder as screen 05 – only now it is live: REC, a timer big enough to read from the wall, a “who's in frame” chip you can change with one touch, and a round stop button. One control, zero decisions mid-climb.
Mark a send or a fall, then choose: analyse now, or move to the next problem.
Shown only when the system cannot decide by itself – one honest question beats a guess that quietly corrupts everything downstream. Analysis runs in the background either way and waits in history, so there is no separate waiting screen anywhere in the flow.
Opens from See the breakdown: the clip, what happened, what to try, the metrics.
The observation, the adjustment for the next attempt, and a metric row (moves, time on the wall, XP) that gives numbers without retelling the story. The primary action is the step-by-step beta – deliberately not called “Analyse”, because that word promises a wait that does not exist.
A send raises a celebratory popover over screen 07, with a teaser of the climb.
The other branch of “How did it go?”, in exactly the same pattern as the fall – a layer over where you tapped, not a new place. The teaser carries all three rewards at once: ability, belonging and commitment. “The 21st climber to send it”, not “rank 21”.
Report today's challenge honestly, and see where it accumulates.
Silent feet is a training game climbers already police themselves, and here it stays that way. The phone does not measure and does not judge – you heard your own feet, you report. The streak line says out loud that rest days do not break it, because a daily streak would punish rest. The XP ceiling is visible so there is something to aim at.
What the session did to your climbing, and what it did to your body – reported separately, because they are not the same measurement.
See what the session did, and decide what comes next.
What changed in your climbing first; points second. The three axes report separately because they are not the same thing. Two closing actions, both facing forward: recover from this session, and set a goal for the next. That is what turns a report into a loop. And one vitals line for watch owners – calories and peak heart rate for the whole session, linking to Body. The same pattern as the popover after an attempt: bottom line here, depth in one place.
Check what the session cost your body, with no watch on.
Readiness fell from 67 to 38. In this tab a fall and a send weigh exactly the same, which is physiologically correct – the forearm does not know whether you topped out. The load map goes from its resting state to loaded: red at the shoulders and arms, amber at the forearms and core. One figure, two states. The figure itself is the control – tapping it opens the full readings, because physiological metrics belong to the body, not to a dashboard card.
Look at the two numbers and understand that they are not the same thing.
The chess.com rule: rating accrues only through play and cannot be bought. XP has a balance, not a level, precisely so it never reads as a ladder. Every card explains its own rules, because a scoring system nobody trusts is worse than no scoring at all. The hexagon carries exactly one teal accent, on the weakest axis.
The identical screen, for someone wearing a watch.
The locked row opens to a real reading and readiness sharpens to 41 – but the tilde stays, because readiness is computed from measurements, it is not itself measured. Two data fidelities in one layout: the difference is a ring style and a tilde, not a different screen.
See six months of progress even when your hardest grade has not moved.
A chart of your highest grade is a flat line, and a flat line discourages exactly the person who is stuck – the product's persona. The pyramid shows the base widening instead. 1,604 attempts are on the board, not only the 148 that worked.
Four sources of challenge, beta from the setter and from climbers, and a league ranked by improvement rather than ability.
DB
Pick a challenge: daily, yours, a friend's, or the gym's.
Four sources, like a daily puzzle in chess. A weekly attempt count makes failure pay directly – today's 31 falls advanced it further than three clean sends would have. Self-reported challenges stay in personal territory and never feed the leaderboards, where only what the gym can verify is counted.
IK
DB
MS
SK
Watch how other people sent it: the setter first, then climbers – fastest first.
The setter films themselves finishing the set, and that clip ships with the problem on day one, which kills the cold start. The second sort, “closest to my build”, exists because the fastest send here belongs to someone 12 cm taller and their sequence will not work for you. Height is captured at setup, so it costs nothing.
D
A
N
See what is happening in your gym, and where you sit in the league.
12th of 96 in your gym moves something; eighty-four-thousandth in the world is noise. Leagues map onto competitions the gym already runs, and they carry no currency of their own – the league ranks by how much your rating rose inside the window, not by the rating itself. Otherwise the strongest climber wins every time and nobody else has a reason to enter.
Understand why a fresh problem has no rating yet, and what that means for your send.
Gyms strip and reset every four to eight weeks, so “new problem” is the normal state, not an edge case. The rating starts from the setter's estimate and corrects itself retroactively as sends come in. The setter's beta is already here, so even a one-day-old problem has something to show.
DP
N
YR
AL
Compare yourself with people who are actually like you.
Four boards, each answering a different question. The default is your own range, ranked by improvement – which is why Dana sits first at 1352 and Yael third at 1498, with you second at 1436. That is not ability order, and that is the point. The world's best sit behind a link that says outright it is not a ranking.
Fifteen seconds of check-in, then a plan for now, tonight and tomorrow. Rest is prescribed here, so nothing elsewhere may punish it.
Answer four questions before you start – about fifteen seconds.
All chips: no typing, no search. Food asks when, never how much – an easier question and a more useful one. Skin is here because climbers genuinely track it, and it is often what ends a session. Pain feeds the injury map; sleep and skin come back in the recovery plan.
Get what to do now, tonight and tomorrow.
The load map finally does work instead of being looked at: the muscles that loaded are the ones that choose the exercises. Food says “within two hours” – no number and no counting. Sleep and skin return from the morning check-in, so all four questions are answered somewhere.
Get told that three heavy weeks have stacked up, and decide whether to back off.
A real training concept, detectable from data the product already collects. The last line is the whole point: a deload still earns XP and still advances challenges. A system that quietly punishes rest teaches climbers to ignore the very warning it just issued.
Mark an area that hurts, and track it over time.
Climbers already track fingers and elbows in their heads. Once the area is marked on a figure the product already owns, the system can warn when load returns to it – for the cost of exactly one new interaction.
Sorted by what you need in the next ten minutes, and a home for the gamification so it stops competing with your identity.
Buy what you are missing right now, sorted by how fast you can hold it.
Sorted by availability rather than category, because you notice the chalk has run out while you are still at the gym. The shoe note counts sessions and points at a resole – the cheap answer. An earlier version derived the same note from a footwork score and pointed at a ₪720 purchase. That version was killed.
Check whether you can try the shoe on today, here.
Local stock outranks cheaper shipping, because a fitting is worth more than forty shekels. Every recommendation states why it is showing – and here the reason is only that you opened the page. Your rating is never used to sell you anything; wear is counted in sessions, not in scores.
N
Pick a season avatar, see badges and achievements, and spend XP.
This is where gamification lives – not next to your name. The photo is identity; the avatar is the game piece that represents you on the gym board. Putting both in one row is two identities competing. Locked items show in grey: you see what you are missing, which is the engine of a cosmetics store.
An empty state should show the shape of the thing that will fill it. The layout never changes between first run and steady state; only the contents arrive.
A new climber sees the shape of what will fill, and what to do first: scan.
Exactly the layout of screen 01, with the content simply not arrived yet. The ring is drawn and unfilled, the chart waits for its first line, the session card holds the place of the first clip. The goal is the only filled element, and the only thing asking to be tapped after Scan.
See the shape that every send will start to fill.
Four hollow rows instead of an empty chart or an apologetic sentence. Seeing the shape – wide base, narrow peak – explains the idea before the first send has happened.
Understand who will appear here, and what it takes to get in.
Three ghost rows instead of a blank board. Entering the board is the same act as earning a rating – three sends – so the empty state points exactly where every other screen points: scan and climb.
See the map, and understand that marking is a single tap.
The figure shows even when there is nothing to show on it, because the figure is the explanation of the feature. One dashed row says what happens when you mark, instead of a screen apologising for having no data.
Even a problem one morning old has something to watch: the setter's clip is on its way.
The setter-beta decision means this list is almost never truly empty. What stays empty is the climbers' shelf – and the dashed row turns that into an invitation: send it and publish first, and you are the one at the top.
Teal is interaction.The one rule that made thirty-four screens read as one product
Not data. Not celebration. Not decoration.
The visual system is small on purpose. An earlier pass had teal doing four jobs at once – interaction, data, celebration and decoration – plus fourteen off-palette greys, two golds and two greens. Cutting it back to one of each is what made a set of thirty-four screens read as a single product rather than a mood board.
Eighteen avatars, twelve badges and seven tokens. Every avatar sits on an identical 78 × 78 canvas, so a single CSS height renders them all at matching optical size – the one with the top hat is a decorated shoe, not a bigger shoe.
Not a restyle. In each pair the desktop version answers the question correctly and arrives too late, and the app version answers a smaller question at the moment it is actually asked.
DB
Until now nobody in the gym had ever watched their own climbing. The attempt is filmed, and the move that failed is frozen and drawn on the body.
Not “try again”, but one concrete thing for the next attempt – next to the beta of someone who sent this problem at their height.
A fall moves readiness, advances challenges and earns XP – and never lowers the rating. The product measures what a climber actually does, not only what worked.
The hardest part of this project was not the design. It was admitting that I was not designing a complex system – I was designing a climbing wall. As long as I thought of it as a system, every feature justified itself and the hierarchy only grew. The moment I looked at it through a climber standing in front of a problem, it became obvious what the core feature was and what had simply been there.
And taking a hierarchy apart is harder than building one. Building is adding. Taking apart is deciding what does not go in, and then explaining to myself why.
I also learned not to accept a direction just because someone experienced gave it. The first version was built that way – and that is what took me the longest to undo.
What I take to the next project: a user's real pain is not what they ask for, and not what is convenient for me to solve. It is specific – it happens in a particular place, at a particular moment. Until I understood it, every other decision was an organised guess.
Out of scope, stated openly: setter tools, gym signage and its legal review, and the seller side of the store.
