How PhysicsWallah moved from private Claude setups to one shared context.
Every analyst had a private Claude setup that only knew the slice of the warehouse they worked in, and PMs waited hours or days for every number. Now every team at PhysicsWallah works from one shared context in Skoot.
The last two months
7x
growth in usage
10,000+
questions asked
1,500+
agent conversations
90%+
of users active every week
- Industry
- Edtech, publicly listed
- Company size
- 10,000+ employees
- Users
- Millions of learners
- Teams on Skoot
- 12+ across 20+ business lines
- Customer since
- July 2026
Challenge
Context lived in private Claude setups
Everyone at PhysicsWallah had Claude. Each analyst built their own projects, context .md files and warehouse connections, and each setup only knew the slice of the warehouse its owner worked in. None of it was shared, so knowledge stayed siloed with whoever built it.
“Everyone has their own Claude projects, but each one stays in that person’s space. There is no common space.”
Bharadwaj Mulakala
Analytics Lead, PhysicsWallah
PMs had Claude setups too, but each one knew only what its owner had managed to piece together, and nothing about other products or business lines. So the real questions still went to an analyst, and every follow-up took hours, sometimes days.
Solution
One shared context for every team
The pilot started with a few analysts on the Pi OTT team. Skoot took in what they already had: their Claude .md files, master queries and data dictionaries. Because it began with the team’s own work instead of a blank page, it gave its first useful answers in the very first week.
Then other teams joined, one at a time. Each team added its own context, and each could use everything the others had added. A PM in one team gets the verified context from another team, without asking anyone.
“It knows the tables better than I do, because it has the verified context from the other teams too.”
Trilok Yadav
Product Manager, PhysicsWallah
Results
Shared context paid off in ways nobody planned
The goal was faster answers. Once every team worked from the same context, it turned out to be useful in ways nobody had planned for.
Root causes in minutes
When Pi OTT subscriptions dropped on 21 June, one prompt was enough. Skoot ran 25+ explorations on its own: it checked payment failures and the Pi widget on the home screen, and compared traffic on Pi OTT against traffic on the main app. That last comparison was only possible because Skoot had context from beyond the Pi OTT team. It named the root cause in under four minutes.
On 27 July, app DAU spiked and nobody knew why. A PM asked Skoot. It split users by exam category, read the funnel and engagement events, and traced the extra traffic to popups and banners. Two follow-ups and under 10 minutes later, the PM had the cause, without pulling in an analyst.
Recurring work runs on Skoot
In most data teams, reports and dashboards take up a big part of every week, and the same work comes back the next week. At PhysicsWallah, that work now runs on Skoot. The weekly Pi OTT report takes about an hour instead of 5 to 6 hours by hand, and dashboards for product experiments ship in 2 days instead of 7.
Analysts stopped waiting on each other
An analyst on one team no longer waits for someone on another team to explain their tables and definitions. That context is already in Skoot, verified by the people who know it best.
Claude stayed, with Skoot inside it
Nobody had to give up Claude. Some analysts and PMs now call Skoot from inside Claude whenever the work touches data, and get the shared context without leaving the tool they already use.