Beyond case studies — I built this ↓
2×
APM Candidate · Delhi, India
SAKSHI
NEGI
2× PM School Winner
I find the real problem before building the solution.
Four case studies. Two competition wins. One honest attempt at thinking like a PM.
Scroll to work
Case Studies
01 / 04
01
Fitso by Cult.fit "How do you make someone come back after Day 1?"
Retention · Gamification · GTM 2nd Place
→
02
VWO Testing "One metric. Four features. Six months."
B2B SaaS · NSM Decomposition · Roadmap Assignment
→
03
Mastodon "70,000 signups in a day. Then silence."
Onboarding · FTUX · Growth · Decentralised Winner
→
04
ShareChat "180M users. A chat feature nobody used."
Adoption · User Research · India · Chatrooms Case Study
→
Improving retention on Fitso
The problem
Users book their first session, show up, then never return. The app had no habit loop — nothing to bring someone back the next day.
Key decision
Scoped to swimming first. One sport, fully validated, before expanding. Reduces eng scope and gives clean signal on whether XP changes behaviour.
NSM chosen
Sessions booked/week — not DAU. In a sports app, passive opens are meaningless. Only showing up counts.
View Full Case Study →
[ Add what you got wrong here ]
6-month adoption plan for VWO Testing
NSM breakdown
Total tests run. Decomposed into experimentation maturity × value realization — two axes that explain why users aren't running tests.
Focus segment
Users yet to experience value — not new users. Already onboarded, familiar with the tool. Highest short-term retention upside.
Top feature
Report Insights — ICE 120. Auto-surfaces winning segments. Removes post-test analysis paralysis that stops users from running the next test.
View Full Case Study →
[ Add what you got wrong here ]
Redesigning onboarding on Mastodon
The problem
70K signups in one day post-Twitter acquisition. Then a cliff. Users couldn't understand what a "server" was or what to do after signing up.
Core insight
Two drop-off points: choice paralysis at server selection and blank screens post-signup. Familiarity bias from Twitter made Mastodon's flow feel broken.
Solutions
Server suggestions by language + interest. Empty states, onboarding checklist, hashtag nudges. Push users to the Aha moment faster.
View Full Case Study →
[ Add what you got wrong here ]
Chat feature adoption on ShareChat
The gap
180M MAU but chatrooms were only used by old-time users who learned features through friends — not through the product itself.
Research
3 personas: Abhishek (connection), Raj (shayari community), Amar (interests/armed forces). Fake profiles + irrelevant content = biggest pain points.
Features proposed
Tooltip onboarding tour, chatroom categorisation to fix discovery, interest-matched rooms surfaced on first visit.
View Full Case Study →
[ Add what you got wrong here ]
Product I built
Subsafe
Know before you are charged
NFLX*7829IND. ₹649.
You know it's probably Netflix.
You don't know which mandate
to cancel, on which app,
through which bank.
Tap what
you pay for.
every month. without asking.
₹0
This is how much left your account.
tap to continue
still charging. you forgot it existed.
74 days since
you opened Canva.
SubSafe caught it before the charge fired
This is how it happens. One forgotten subscription at a time.
tap to continue
the invisible charge
Money left.
You don't know
who took it.
HDFCBK · bank alert
Rs.649.00 debited from a/c XX4821 on 12-Jan by NFLX*7829IND. Avl bal Rs.18,432.
9:14 AM
Who is NFLX*7829IND?
Was this my card, my UPI, or my wallet?
Can I stop it — and how?
charge identified
merchant
Netflix
charged via
UPI autopay · a/c XX4821
annual cost
₹649 × 12 = ₹7,788
next charge
12 Feb · 31 days away
tap to continue
No surprise autopays.
No ₹7,788 shock
at year end.

SubSafe
Know before you are charged
What I built
and why
01 / 03
what surprised me
Most people didn't know subscription trackers existed. The apps that did exist gave them a prettier list — same problem, different screen.
A list doesn't create urgency. ₹7,788/year does.
02 / 03
what challenged me
SubSafe needs bank SMS access to work. That's the most sensitive ask on an Indian user's phone.
So the ask comes after the first subscription is in — not at onboarding. Trust first, permission second.
03 / 03
what I'm still figuring out
The people I was building for were already exhausted by subscriptions. I was about to add one more.
SubSafe would show up in its own subscription list. I priced it so it's the first one worth keeping.