Pratham Prasoon
Independent UX PracticeEst. 2024
Portfolio Vol. 1 · Spring 2026

Designing
interfaces
that get out
of the way.

I'm a designer working at the intersection of cognitive psychology and software craft. This volume gathers three case studies from my Human-Computer Interaction studies under Dr. Samit Bhattacharya at IIT Guwahati — each one an attempt to apply Norman, Nielsen, Fitts, and a lifetime of Gestalt instinct to interfaces real people will actually live inside.

06.103.115
Norman · Nielsen · Fitts · Hick · Gestalt · Shneiderman · Card & Moran · WCAG 2.2 · Norman · Nielsen · Fitts · Hick · Gestalt · Shneiderman · Card & Moran · WCAG 2.2 ·
A · Manifesto

A short list of the things I believe.

“The user is never wrong. If they cannot find the button, the button is hidden — even if it is in plain sight. Affordance is not a property of the object; it is a relationship between the object and the person.” — after D. Norman, The Design of Everyday Things, 1988
i.

Subtract before adding. Every pixel competes with the one beside it. Heuristic 8 is, at heart, a moral position.

ii.

Predict before testing. KLM costs nothing and surfaces 70% of the friction a usability lab would have shown you next month.

iii.

Design for the worst day. Low light, slow network, an anxious user, a cracked screen. If it works there, it works everywhere.

01

Selected work.

Three case studies
2025 — 2026
Case 01 / BankingDharma
A redesign of a public-sector banking app for first-time smartphone users in Tier-3 cities. The brief: bring transfer success rate from 41% to above 80% without increasing the number of fields on the form.
2026 · 14 weeks
solo · UX lead

The problem

Field interviews in Patna and Indore surfaced a single recurring failure mode: users filling the IFSC field with their own account number — because the labels above the two inputs were visually identical, and the icons (a tiny bank vs a tiny card) too small to disambiguate at arm's length on a 5″ phone in midday glare.

The existing app violated heuristic 5 (error prevention), heuristic 6 (recognition over recall), and a quietly devastating amount of Gestalt similarity — the page read as one undifferentiated block.

Persona

M
Meera, 47
Tailor · first smartphone (2024) · Patna
“I am afraid to press the wrong button. Last time I sent ₹500 to a number I did not know.”
Goal: send money to her son in Bangalore

Approach

The redesign was driven by three commitments. First — apply Fitts's Law to the most common action (transfer): a 96px primary tap target at the bottom-third of the screen, where the thumb already lives. Second — break the form into one-question-per-screen (Hick's Law: reduce choice density). Third — at every confirm step, restate the action in plain Hindi-English mixed copy, with the recipient's name and amount in Newsreader 28pt italic so the eye has something to anchor on.

Heuristic evaluation (5 reviewers)

#HeuristicBeforeAfter
5Error prevention3.8 / 40.4 / 4
6Recognition vs recall3.2 / 40.6 / 4
8Aesthetic / minimalist2.6 / 40.8 / 4
9Recover from errors3.0 / 41.0 / 4
scores: severity, 0=none → 4=catastrophic
— before · cluttered single screen
DharmaBank · Quick Transfer
FROM ACCOUNT
XXXXXX4729 — Savings
PAYEE A/C NO
PAYEE NAME
IFSC CODE
AMOUNT
REMARKS
PURPOSE CODE
CANCEL
CONTINUE
Refer to your passbook for IFSC.
— after · one question per screen
STEP 3 OF 4 · ●●●○
How much
would you like
to send to Aman?
₹2,500
+₹500
+₹1,000
+₹2,000
CONTINUE →
Tap and hold for voice readback
Fitts's Law Hick's Law H5 Error prevention H6 Recognition Gestalt similarity UCD · 14 interviews
+97%
Transfer success
41% → 81%, n=126 trials
−54%
Time to first transfer
3:12 → 1:28 (median)
−78%
Wrong-field errors
eliminated by step-form
4.6
SUS score (out of 5)
up from 2.8
Case 02 / HealthcareSaanjh
A free-standing kiosk for outpatient registration in a rural government hospital. The user has never touched a screen before. The light is brutal. The line behind them is long.
2025 · 9 weeks
with PHC Sahyog

The problem

Outpatient registration at PHC Sahyog took 9–14 minutes per patient through a single paper-based desk. Patients arriving at 6 am were often still waiting at noon. A previous kiosk pilot had been abandoned after staff observed elderly patients “pecking” at the screen with one finger, missing buttons, and walking away in frustration.

The previous kiosk had buttons of 32 × 32 px spaced 8 px apart. Per Fitts's Law on a 1.4m-distant kiosk screen, this is a near-impossible target for an unsteady hand.

Constraints

WCAG 2.2 AAA contrast (this is a hospital, not a brand). Fingertip-grade hit targets (≥ 72 px). Voice-first prompts in Hindi, Bhojpuri, and English. No more than one decision per screen. No use of the colour red except for true errors.

KLM analysis · before vs after

BEFORE — paper desk
M · find form on rack1.351.35s
P · point to clerk window1.102.45s
K × 38 · clerk types into ledger0.2010.05s
R · wait for ID assignment0.5010.55s
(plus wait time, mean 9–14 min)
TOTAL EXPERT TIME≈ 11s + queue
AFTER — Saanjh kiosk
M · approach kiosk1.351.35s
P · select language card1.102.45s
P × 4 · large symbol-buttons1.106.85s
R · token printed0.507.35s
TOTAL EXPERT TIME7.4s · no queue
— before · 32px buttons, 6 colors, 4 fonts
SAHYOG HOSPITAL — PATIENT REGISTRATION KIOSK SYSTEM v2.4
New
Existing
Emergency
Pharmacy
Lab
Bills
Please select an option from the menu above and follow the on-screen instructions. For assistance, contact the help desk between 9:00 AM and 5:00 PM. *आधार कार्ड लाना अनिवार्य है*
Form A
Form B
Form C
Form D
OPD
IPD
DOTS
Vacc.
— after · 4 cards, 96px+, AAA contrast
SAANJH KIOSK · RURAL HEALTH
Why are you
here today?
First
visit
Coming
back
Emergency
Help me
हिं
EN
▶ tap any card
Fitts's Law (96px) WCAG 2.2 AAA H2 Real-world match H4 Consistency Gestalt proximity Voice + visual
7.4s
Expert KLM time
predicted before pilot
12:1
AAA contrast ratio
ink on bone
96px
Min hit target
3× iOS guideline
0
Misclicks observed
n=42 first-time users
Case 03 / EducationPravah
A complete information-architecture rebuild of an NPTEL-style course dashboard. The brief was small: “make it less stressful.” The work, predictably, was not.
2026 · 6 weeks
solo · IA + UI

The problem

A learner opens the dashboard and sees: 14 tabs, 3 carousels, 2 banners, an assignment countdown, a discussion forum widget, and a “Recommended for you” strip that recommends the same six courses every week. There are too many entry points and no clear mental model of where I am, what I owe, and what's next.

This is a textbook violation of heuristic 1 (visibility of system status), of Miller's 7 ± 2, and most painfully — of the user's actual goal, which is to study.

Card sort (n=18)

18 learners sorted 34 dashboard objects into self-named groups. The convergent clustering collapsed cleanly into three buckets: “due now”, “keep learning”, and “explore later”. The redesign uses exactly that.

Information architecture

─ Pravah · home
  ├── due now (time-sensitive only)
  │    ├── assignment · WK6 · 2d
  │    └── quiz · WK5 · 5h
  ├── keep learning (in-progress)
  │    └── HCI · L12 of 32
  └── explore later (collapsed)
       ├── catalog
       ├── forums
       └── certificates
tree depth: 2 · top-level breadth: 3

Cognitive walkthrough

For the goal “submit my Week-6 assignment,” the previous design required 5 decisions across 3 pages. The redesign requires 1 decision (tap “due now” item) and 1 confirmation. Steps fell from 5 → 2; M-operators from 3 → 1.

— before · 14 tabs, 3 carousels, no hierarchy
Home
My Courses
Catalog
Calendar
Grades
Discuss
Notes
Resources
Wiki
Help
Profile
Logout
🎁 NEW! Take 20% off our Premium Plan — Limited Time!
Course A
Course B
Course C
Course D
⚠ 3 assignments due
✓ Streak: 5 days
★ Earn badges
Recommended carousel ▸ ▸ ▸ Trending ▸ ▸ ▸ Continue ▸ ▸ ▸ Forums ▸ ▸ ▸
— after · three buckets, in priority order
Pravah
MON · 19 APR
DUE NOW
Assignment 6 — in 2d 14h
HCI · Cognitive Walkthrough
KEEP LEARNING
Lecture 12 · 38% complete
+ explore later (3)
Information architecture Card sort · 18 users H1 Visibility of status H7 Flexibility H8 Minimalism Miller 7 ± 2
5→2
Decisions to submit
cognitive walkthrough
−71%
Time to “due” item
5.8s → 1.7s, n=22
3
Top-level objects
down from 14
+38
SUS points
42 → 80
02

A user-centred process.

After ISO 9241-210
+ Norman + Cooper
i.
Understand

Contextual inquiry. Field interviews. Ride-alongs. Speak less, watch more. Notice what users don't say.

ii.
Specify

Personas with goals (not demographics). Task analysis. Hierarchical Task Analysis when the work is procedural.

iii.
Predict

KLM, GOMS, cognitive walkthrough. Heuristic eval against Nielsen. Cheap, fast, before any pixel.

iv.
Prototype

Paper → Figma → coded. Each step exists to kill the wrong ideas earlier.

v.
Evaluate

5-user usability tests, think-aloud protocol. Severity ratings. SUS for the file, not the wall.

vi.
Iterate

Then again. Then again. The first version of any interface is wrong; the question is how cheaply you find out.

03

The toolkit.

Principles applied
across the work
D. Norman · 1988
Affordance
Objects suggest their use. A flat plate affords pushing; a handle, pulling. Don't fight the suggestion — design it.
D. Norman
Mapping
Spatial correspondence between control and effect. Knobs in the same shape as the burners they control.
D. Norman
Feedback
Within 100ms — perceived as instant. Within 1s — flow uninterrupted. After 10s — user mentally leaves.
D. Norman
Constraints
Make the wrong action impossible. Stronger than any error message you can write.
J. Nielsen · 1994
10 Heuristics
Visibility, match, freedom, consistency, prevention, recognition, flexibility, minimalism, errors, help.
B. Shneiderman
8 Golden Rules
Strive for consistency. Cater to universal usability. Offer informative feedback. Design dialogs to yield closure.
P. Fitts · 1954
Fitts's Law
MT = a + b·log₂(D/W + 1). Big, close targets win. Always. Especially under stress.
W. Hick · 1952
Hick's Law
RT = a + b·log₂(n + 1). Choices have a cognitive price. Group, hide, tier — don't flatten.
Wertheimer et al.
Gestalt
Proximity, similarity, closure, continuity, figure/ground. Perception arrives before reading does.
G. Miller · 1956
7 ± 2
Working memory holds about seven chunks. Honour the limit. Or chunk smarter.
Card · Moran · Newell
KLM / GOMS
Predict expert task time before you build. The cheapest usability test you'll ever run.
W3C · 2023
WCAG 2.2
4.5:1 body contrast. 3:1 large. Perceivable, operable, understandable, robust. Non-negotiable.
04

Said about
the work.

Course feedback
peer reviews · 2025–26
What stayed with me was the restraint. Most students decorate; this is the only portfolio in the cohort that subtracts.
— Course peer review, Studio II
The Saanjh kiosk study reads like a piece of applied cognitive science. The KLM tables alone would justify the grade.
— TA feedback, HCI 06.103.115
Rare to see Norman cited in service of an actual user, and not as a footnote ornament. Clear, purposeful, and quietly excellent.
— Final critique, Spring 2026