M · 255+ years · 5× a week
- Tracks everything in a logging app
- Rests by pulse and breathing, no timer
- Left apps when the best features moved behind a paywall
“Metrics interest me least.”
UX / UI Design Case Study
A strength training app built to help you make sense of your progress.
Nataliia Lyfarenko · 2026 · Concept project
Project overview
A strength tracker that explains your progress
ProblemTrackers log workouts and leave users alone with a table of numbers.
→
SolutionEvery screen interprets the data: recovery map, breakdowns, personal correlations.
14
survey respondents
4
interviews
6
test participants
6
weeks, hypothesis to test
- Role
- Research, UX, UI and design system, solo. A medical doctor reviewed the health screens.
- Tools
- Figma, Google Stitch for AI drafts, Claude Code, survey, interviews.
- Context
- Concept project: both flows are designed and prototyped end to end.
Research · Survey and interviews
A problem confirmed by survey data
Survey · n = 14 · highest first
64.3%
don't know which muscles to train today
9 / 14
57.1%
can't tell whether they are progressing
8 / 14
57.1%
lose motivation when the numbers show no result
8 / 14
42.9%
don't know what to do with the charts
6 / 14
Asked for
92.9%
short personalization quiz at first launch
85.7%
rest-end alert outside the app
71.4%
muscle map on the home screen
71.4%
comparison against real achievements
Interviews · n = 4
F · 301.5 years · 3× a week
- Cannot tell if she is progressing, only weight counts
- Feels stronger or weaker by cycle phase
- Dislikes feeds full of strangers
“Statistics for workouts by cycle day: I haven't seen that anywhere, and it interests me.”
F · 251+ year · 3× a week
- Writes the program in notes, because the app has none
- Wants chart legends she can reopen
- Wants to hide pulse, kcal and weights from friends
“I try not to judge all my progress by such days.”
F · 26About a year
- Quit a logging app: entering sets took too long
- Rests by the gym's red-green light
- Skips the gym on painful cycle days
“I would like to always read the workout conclusions.”
Own PUMPT survey, April 2026, n = 14 · written interviews, participants anonymized
Three findings, three design decisions
ProblemThe app logs the workout but leaves its result unexplained
57.1%
can't tell if they are progressing
Users have the data and can't read it.
All three women interviewed say cycle phase and sleep change what they can lift.
Design decision
Session summary
Every card compares today with your own recent sessions.
85.7%
at least sometimes skip logging
Slow logging breaks the habit.
One interviewee quit a logging app because entering sets took too long.
Design decision
Fast logging
Weight, reps and RPE in one row, with an optional rest timer.
78.6%
accept social features if optional
Social features work as an option.
Two interviewees dislike feeds full of strangers.
Design decision
Friends, field by field
No feed, and a switch for every field you may share.
PrinciplesCalculation over guessworkMinimum tapsSocial by choiceData as hero
Competitive Analysis
Where market leaders fall short
HEVY10M+ · Featured by Apple
FITBODAI-powered · Subscription
STRONGBasic logbook · PRO paywall
MACROFACTORAdvanced · Science-based
PUMPT
Data visible on the home screen
Hidden
Behind a paywall
Behind PRO
Overloaded
Yes
Rest timer
—
Off
—
—
Opt-in
Onboarding by experience level
Generic
Overwhelming
No
Advanced users only
7 steps
Also found
- No home dashboard, progress is buried in Profile
- Social feed takes priority over workout numbers
- Paywall on the very first screen: €99.99/year before the first workout
- Empty dashboard with zero guidance for new users
- Dashboard and analytics fully locked behind PRO
- Zero workouts and a blank template, no onboarding at all
- Periodization and RIR overwhelm beginners from day one
- Too much data crammed into the workout screen at once
TakeawayEvery app has a strong side, and none solves all four problems at once. PUMPT is built to solve all four.
Original competitive analysis, 4 testing sessions, 2026
Define · Personas, journey, flows
Kate and Maria’s training week
Built from interviews P2 to P4. The thinking row is my reading of what the participants said.
KATE25–30 · 1 to 1.5 years of training
Cannot tell whether she is progressing on days shaped by cycle phase and sleep, and keeps exercise notes outside the app.
ALEX25 · 5+ years of training
Extra metrics add noise.
MARIA26 · beginner, about a year of training
Entering sets took so long that she stopped logging and keeps everything in memory.
1Plan
2Train
3Log
4Review
5Share and recover
Doing
Builds the week in notes, tests it, corrects it
Picks exercises by muscle group; for a new machine borrows the weight of a similar one or scans the QR code on it for a video
Logs sets and RPE in the app, writes technique somewhere else
Compares weight and reps over time
Shares with friends; rests by breathing, pulse or the red-green light in the gym
Thinking
Will this program work for my week?
Is this weight right for a machine I have not used?
Where do I write how this set felt?
Is this a bad week or one bad day?
Who sees my pulse and calories?
Pain
The program lives in a notes app
No weight guidance for new machines; a busy machine breaks the plan
No place for tempo, pauses and personal technique; entering sets takes so long that some stop logging
Cycle, sleep and stress distort single days; chart legends appear once
Pulse, calories and weights cannot be hidden one by one
Opportunity
Routines and notes pinned to the exercise (Routine detail, Exercise notes)
Swap exercise with a suggested weight (Swap sheet)
RPE in the table; set details for set type and failed set
Workout breakdown, cycle context, legends on request
Follow model with visibility per field (Sharing settings)
Kate combines interviews P2 and P3. Alex is based on P1. Maria is based on P4 and the survey profile (57% train under six months). No demographic details were invented.
Two flows that carry the findings
Flow A covers onboarding with a female profile and cycle context. Flow B covers logging, adapting and explaining a workout.
Flow A · Onboarding with a female profile and cycle metrics
SplashWelcomeAccountPrivacyNameAge checkGenderCycle trackingCycle situationsGoalExperienceEquipmentWeightHeightHealth syncCycle data consentNotificationsSetup complete
Flow B · Home, workout and metrics
HomeHome with cycleCycle insightsWeekly recapRoutinesRoutine detailCreate routineExercise libraryActive sessionEnter weight, reps, RPESet details: set type, failed setExercise notesSwap exerciseAdd exerciseStop your workoutSession summaryWorkout breakdownHow do you feelSharing settings
System screens and states
Log inForgot passwordEmpty statesOfflineLoadingConfirmationsSettingsHealth and dataAbout and legalPro plan
Log a failed set3 taps
Tap the set→Failed set→Save
Swap a busy machine3 taps
Exercise menu→Pick an alternative→Confirm
Add a technique note3 taps
Tap Notes→Write→Save
Screens · Flow A
Onboarding and cycle context
Personalization from the first tap
Name
Gender
Goal
Experience
Equipment
Weight
Height
92.9% of respondents want a short personalization quiz at first launch
Cycle context, built on consent
Off until an explicit yes. Once on, it runs through Home and has its own section.
What the evidence saysDesign rule in PUMPT
No universal phase effect on strength or RPENo phase-based training tips
Large individual differencesPatterns only from her own data
Small effects, noisy dataA minimum of data before any pattern; “may”, “so far”
A pattern may not exist“No clear pattern” is a valid result
Symptoms matter more than phaseOptional symptom and pain logging, always private
Reviews find no universal effect of cycle phase on strength: McNulty 2020, Blagrove 2020, Niering 2024.
Screens · Flow B
Workout, progress and sharing
In view exactly when it matters
78.6%
find the rest timer inconvenient
85.7%
want the end-of-rest alert
3/4
interviewees rest by breath, pulse or the gym light
Progress is visible at a glance
71.4% asked for the muscle map on Home · 57.1% asked for the next workout in view
Only the friends you choose
78.6%
accept social features if optional
Follow by usernameNo requests, no leaderboard
Visibility per fieldPulse, kcal and weights can stay private
Health & AI safety
Designed with a medical doctor
Cycle, safety and AI screens were reviewed by a Health & AI Safety Lead: an M.D. who also evaluates LLMs for reasoning reliability and crisis-response safety.
The principle every screen follows
Describe the data. Never diagnose. Never prescribe.
Two levels of safety
A calm flag when the data is worth a conversation. One instruction when someone feels unwell.
She starts it herselfWorkout data can’t reliably show an emergency, so the app doesn’t guess.
Pain stays a note“Pain or discomfort” marks where it hurts. No diagnosis, and the note is always private.
The streak can pauseWeeks of met goals can pause without loss, so nothing pushes her to train through pain.
What the review changed
What the review covered
- Evidence review on cycle and strength
- Clinical review of the screens
- Red-flag logic and escalation
- Rules for the AI recap
What changed after it
- Estimated phases and hedged wording
- Special cycle situations
- Recovery categories instead of “88% recovered”
- A two-level safety flow
- Disclaimer, age check, symptoms always private
What the LLM sees
- Weekly figures computed in code, exercise names
- Cycle day or phase only with separate consent
- Never: name, notes, raw Health data, weight, friends
- Nothing at all when “Personalize with AI” is off
Allowed
“You haven’t logged a period in about 3 months. Changes like this can have many causes and are worth discussing with a doctor.”
Not allowed
“This may indicate RED-S.”
“You’re in your luteal phase: go lighter today.”
A concept case. A real launch also needs a regulatory assessment, a lawyer and a DPO for consent and GDPR health data, and a licensed physician’s final review of flags and escalation texts.
Process · Evolution and testing
The active session, redrawn three times
UnderstandSurvey · competitors · interviews
DefineFindings · personas · journey · flows
PrototypeStitch drafts · Figma · two flows
IterateDesign system · themes · accessibility
What the first testers showed
Six people on their own devices: 11 tasks with an ease rating after each, the System Usability Scale, and an A/B comparison of the session summary.
6
participants: 4 usability test sessions, 4 A/B sessions
87.5
mean SUS from 3 valid questionnaires, 77.5 to 95
4 / 4
started a workout in one tap, median 5.7 s
4 / 4
finished onboarding with the cycle block, median 108 s
A/B: session summary, 4 sessions, both versions
4 / 4
found what improved on the new screen, median 2.6 s
4 / 4
preferred the new screen, whichever they saw first
“In the first version what you need stands out more; you don’t have to search for it.”
A participant who saw the new screen first
Changed after testing
RPE moved into the tableTesters looked for RPE where they already enter weight, so it became a column next to weight and reps, entered the same way. The set number opens the rest: set type, failed set, pain.
Changed after testing
The pattern answers firstMy patterns now opens with the answer in one line, and the charts sit below as evidence. On Home the card is named “Your cycle pattern” and shows that it opens.
Iterate · Design system
One token set, two themes
Each measured property keeps one colour on every screen, so the colour says what is measured before you read the label.
Tokens · dark → light
bg-base→#11131B / #F5F6FA
bg-surface→#1C1C28 / #FFFFFF
text→#F0F0F8 / #14151F
accent text→#5B8AFF / #1F52D6
accent→#2D6AFF / #2D6AFF
error→#FFB4AB / #B3261E
16.4 : 1
primary text, dark
16.8 : 1
primary text, light
≥ 5.1 : 1
tertiary text, both
Bebas Neue for display 48 / 32 / 24 · DM Sans for text 18 / 17 / 15 / 13 / 11
Colours with meaning
- Volume and setsBars against recent sessions · #5B8AFF
- EffortRPE on a 1–10 scale, failed sets · #FF9F43
- TimeA ring against the planned duration · #4FC3F7
- PulseHeart rate across the session · #FF6B6B
- RecordsEstimated 1RM trend, stars · #F5C542
- RecoveryFresh muscles on the body map · #45C8A0
- CycleCycle day, estimated phase, patterns · #C084FC
- GrowthOnly “better than last time” · #3DCB8A
Components
Prototype · Two flows
Clickable prototype
Tap the screen the way you would on a phone. Both flows run right here, from the first screen to the last.
Screen
Tapping outside an active area highlights where you can tap.
Outcomes
What I learned
The survey and the interviews together showed that the rest timer should be optional, and the product follows that.
I drafted screens with AI in minutes, then the survey and the interviews decided what stayed.
Experienced users and beginners need different things, so explanations appear once and reopen on request.
TakeawayI generated quickly with AI, checked every screen against the research and built the system from what held up.
Future features
What I would build next
In progress
Cycle-aware progress insights
Patterns appear once enough cycles are logged; the threshold is under medical review.
Next
Inclusive mode
Color-safe states, larger type, full VoiceOver flows and voice logging.
Program import with AI
A program from notes or a photo becomes routines. One interviewee keeps hers in notes.
Later
Programs by level
Ready-made plans from coaches for each level. 57% of respondents train under six months.
Sharing presets for friend groups
Choose once what each group can see. One interviewee wants to hide pulse, kcal and weights from friends.
Optional friend challenges
Off by default, no public leaderboard. 78.6% accept social features if optional.
