Built a system that grew an exam prep channel from 7,000 to 100,000 subscribers.

Three channels, and not one of them had a system behind it.

SIFF Ventures ran Western Vocals for adult singers, a piano and keyboard program, and Parikshe. Content went out whenever it went out.

I started on Western Vocals, ran the audience research and a long diagnostic on what was actually driving growth there, and found the demand was not big enough to build on. Parikshe was where the audience already was, so that is where I moved.

The strategy came out of the data before it came out of anything else. I went through the channel's YouTube audience data and every other source we had, ran the analysis at category level and then at individual player level, worked out what was already performing and why, and replicated it. Then I A/B tested everything on top of that, thumbnails, titles, formats, hashtags.

What I found early was that nobody running these channels really knew what they were doing, and nothing was built to any standard. Every video was going out with 20 hashtags on it, and hashtags were bringing in 0.0% of the traffic. Cutting them to the three or four that were actually relevant took hashtag traffic up by over 200%. There were a lot of small fixes like that, and together they moved more than any single big idea did.

Then I asked the students directly. The poll came back with 1,401 responses, and I ran a focus group with 30 SSLC students on top of it, most of them already using Parikshe to prep. Cheat sheets came up over and over: quick, condensed reference content to revise from. I built a whole content category around it and it became one of the most-watched things on the channel. Off the same conversations I built separate WhatsApp communities for teachers and for students. Community posts were running around a 12% engagement rate at their best.

Then the machine to run it on:

• A ClickUp-gated upload pipeline, so nothing published before it cleared review.
• A six-step community-post workflow: ideation, day-wise plan, cross-check, design QC, Kannada translation, schedule.
• A comment system that took two to three hours a day down to under 30 minutes, replies drafted in batches and every one read by a human before it went out.
• 15 KPI dashboards, so every format had a number attached to it.
• A biweekly A/B testing cadence, so the next test never waited on someone remembering to run it.

The channel went from 7,000 to over 100,000 subscribers while I was there, and Instagram from 1,000 to more than 50,000 over the same stretch. The YouTube channel has passed 250,000 since, on the systems I documented and left behind, which is the part I did not have to be there for.

The outcome I would point at is not the impressions. It is that the analytics and the content systems underneath them keep learning: every format carries a number, the tests run on a cadence, and the next decision comes out of the last one. That is what kept compounding after I left.

It also gave the company a base to sell from. SIFF went on to sell courses and paid study material off the back of the channel, to an audience that had already told us exactly what it was stuck on.

Parikshe had 7,000 subscribers and no way of knowing what its audience wanted. I left it with 100,000-plus, five documented systems, and a company that had something to sell them.

This is where I started thinking in systems rather than in posts. A single good video is one result. A pipeline that reads its own numbers, tests against them and feeds the answer back into what gets made next keeps producing results with nobody holding it. I came in with content ideas. I left with documented workflows.