Work
Work
Every case study on this site. Featured pieces first, then shorter work.
Featured
Building for people who can write English but freeze when they speak it
OutLoud
An AI speaking coach that gives you a private place to practice out loud. Researched, scoped, built, shipped, and tested with real users on their own phones.
- Live
- BUILT & MEASURED
- PRIMARY RESEARCH
100+ sessions · 30+ devices
Finding out the problem was narrower than it looked
Uber
Pickup friction is usually framed as something every rider faces. Eight interviews showed it concentrates in four venue types - which changes what you'd build first.
- USER RESEARCH
- PRIMARY, n=8
Discovery and research · May 2026
Pressure-testing a strategy against two incumbents' live pricing
Emburse
A three-year product strategy for B2B expense management in India and APAC, with the competitive position checked against what Zoho and Expensify actually ship.
- B2B STRATEGY
- DESK ANALYSIS + VERIFIED
Choosing a metric that can't be gamed by a discount campaign
Zomato
Funnel hypotheses, four retention cohorts, and why Weekly Ordering Households beats order volume as a North Star.
- METRICS & COHORTS
- DESK ANALYSIS
Funnel, cohorts and metrics · Tier-1 city food delivery
Library
A membership plan, an off-season dip, and what I learned about marketing
Salon
Co-founded a unisex salon, found a seasonal revenue problem, built a membership mechanic against it, and took two more years to work out the real constraint was upstream.
- OPERATING BUSINESS
- PRIMARY
Unisex salon · Thane · operating since February 2023
Method
Operating a business, three years of DINGG data
Finding
Off-season revenue improved, but the binding constraint turned out to be marketing, not pricing.
Why price trust, not price, was the AOV lever
Zepto
A strategy exercise on raising average order value in quick commerce, where the useful finding was that the blocker is perceived value, not actual pricing.
- PRODUCT STRATEGY
- PRIMARY, n=11
Survey n=11 · quick commerce, India · average order value
Method
Survey of 11, plus a live competitor check
Finding
Respondents assume the kirana is cheaper and never check - the blocker is price trust, not price.
Activation is the wrong metric, and that is the whole problem
Twitter (X)
An AARRR funnel analysis of X, where the central finding was that the product defines activation as posting while most new users arrive to read.
- GROWTH ANALYSIS
- DESK ANALYSIS
AARRR funnel · new-user journey · acquisition to referral
Method
Product walkthrough against the AARRR funnel
Finding
X defines activation as posting, while most new arrivals come to read.
Turning seven usability findings into three shippable fixes
Strava
A heuristic evaluation of Strava's Record an Activity flow, and the harder part - deciding which findings were worth a sprint and which were just true.
- UX EVALUATION
- DESK ANALYSIS
7 issues · 10 heuristics · 3 UX laws · 3 first-sprint fixes
Method
Heuristic evaluation, 10 heuristics and 3 UX laws
Finding
Nearly every issue lands at the moment the user is least able to read the screen.
Designing a habit loop, not a feature list
VitaFit Together
A PRD for gamification and community features on a fitness platform, where the design decision that mattered was making the two features depend on each other.
- PRODUCT REQUIREMENTS
- FICTIONAL SCENARIO
Gamification and community · first-month retention
Method
PRD and flow design for a fictional scenario
Finding
Streaks and challenges only work paired - and the rest allowance has to ship in V1, not V1.1.
Treating privacy as the constraint, not the trade-off
AltAcad
A systems design proposal for a personalised learning recommendation engine, where the central decision was to cap personalisation at what privacy allows rather than the reverse.
- SYSTEMS DESIGN
- FICTIONAL SCENARIO
Method
Systems design for a fictional scenario
Finding
Privacy is the constraint, not the trade-off - accuracy bought with trust buys nothing.