Shravan’s Round: a mock interview for people switching into product.

Build something for people moving into product from any background. That was the whole brief, and the important word in it is into. Not product managers who want to level up: people who are not product managers yet.

That choice decides almost everything downstream. An experienced PM walking into a mock needs a harder case. A switcher needs something else first. They have done the work that product management is made of, shipping, triaging, forecasting, selling, firefighting, but they have never had to say it in the language an interviewer is listening for. They lose the round in the first four minutes, before a single case is opened.

so why would the practice start at the case?

It does not. Every round opens with the transition story, and the product carries six different ladders for it: engineer or QA, analyst or data, operations or sales, consultant, founder, and student or fresher. The same question lands differently depending on where you are standing, so the follow-ups differ too. The consultant ladder is marked as weakly supported in our own notes, because the source material on consultants is thin. We would rather flag that than fake it.

Shravan Tickoo is a well known name in Indian product management and the founder of Rethink Systems, where I was part of Cohort 8. The people asking him how to break into product outnumber the hours he has by an enormous margin. That gap is the product.

So this was never built to be sold. It is free, there is no login, no card and no form, and there is no paid tier waiting behind the score card. We had designed one and then removed it, because a paywall on the last screen would have turned a gift to the community into a funnel. The people who most need a second practice round are the least likely to pay for one.

A persona prompt can imitate a tone. It cannot answer a follow-up the way a specific teacher would. So the interviewer sits on top of everything Shravan has actually said in public or in class: 18 cohort sessions, 165 long videos and 414 shorts, 597 documents in all. Every one is kept twice, as the raw transcript and as Markdown with its own front matter, so any answer can be traced back to the source it came from.

The interesting decision was what not to clean. Some of that material is a mess. One session is a rough Zoom auto transcript with other cohort members talking over him. One is flagged low confidence because the content may not match its label. Another has its parts in an unverified order. Large stretches are Hinglish, mid sentence.

All of it was kept, flaws marked in the front matter rather than edited out. Tidying the transcripts would have tidied away the thing worth capturing: how he actually talks when he is teaching, interrupting, switching languages and going back on himself. The chunks are tagged by topic and question type on the way in, so a messy source can still be retrieved precisely, and the known defects are recorded where the ingestion script can see them.

Twenty minutes, and most of it is the candidate talking. Two setup choices, then the transition story, then one product sense or design case, then a light probe on metrics. Say something loose and it gets picked up later, because the next question is chosen from what you just said rather than read off a list. There is no script to memorise, which is the point.

The case bank holds twelve cases lifted from mocks Shravan actually ran, things like designing an ATM for a blind user or explaining a ten percent dip in creator follower growth. Indian products, Indian context. The score card rates five dimensions, structure, product sense, metrics, prioritisation and communication, and then names one thing to fix before the next interview. One, not a list, because a switcher leaving with nine weaknesses does not go and practise.

Answering in Hinglish is fine and is transcribed and scored as it comes. Clarity is judged, accent is not.

• 30hostile linesThe persona was tested against thirty deliberately awkward candidate replies, including flattery, panic and refusal to answer.
• 1hard ruleOut of that testing came the rule the persona is never allowed to break: never make the candidate feel small.
• 5cases liveFive of the twelve cases are in the current build. Guesstimates and heavy execution rounds were left out of the MVP on purpose.

I own the brain here, not the whole product. Voice and the avatar are other people’s work on the same codebase, and the honest version of this case study says which parts are mine and which are still bets:

What’s proven: it is live, free and finishes a full round. The interviewer answers follow-ups out of Shravan’s own material rather than out of a generic PM script, and testers came back with interaction complaints, not credibility ones.
What’s designed but unproven: that a rehearsed transition story actually changes the real interview. That needs candidates who sit a round here and then go to a real one, and we do not have that data yet.
What I’d do differently: get the content in front of Shravan sooner. The persona, the rubric and the twelve cases were pulled out of his recordings before he had reviewed any of it, so every early answer rested on my reading of what he meant.