Rishabh
Bahuguna

I was scrolling the podcast section when I noticed a small blue dot on an episode I hadn't opened. I knew what it meant — Spotify had tracked that I'd listened to earlier episodes, flagged this one, and placed that dot precisely to catch my eye at the right scroll depth. I didn't just think "useful feature." I thought: what signal tells the system I am a returning listener worth re-engaging, versus a first-time visitor worth onboarding differently?

That question pulled me into how Spotify's recommendation engine actually works. According to USC's Viterbi School of Engineering, Spotify's system combines three interlocking layers: collaborative filtering that maps relationships across users who share listening habits, NLP models that scan lyrics, blogs, and social media to extract mood and cultural context, and audio analysis that classifies songs by sonic attributes like tempo, energy, and danceability. Discover Weekly drops every Monday because listening intent peaks then. Release Radar lands Friday because new music does. The timing is not aesthetic — it is a retention hypothesis. Every skip, every save, every lingered-on artist page is implicit feedback reshaping what surfaces next. That blue dot wasn't a design choice. It was a product decision backed by a user behaviour model. That is the level I want to operate at.

The streak never moved me as much as the Friend Quest did. The moment a friend's progress appeared next to mine, I opened the app not out of habit — but out of genuine accountability. That is a different user state entirely. And it made me wonder: how many of Duolingo's daily actives are showing up to learn, and how many are showing up so they don't lose?

Jackson Shuttleworth, Duolingo's Group PM for Retention, confirmed in a public interview that the team ran over 600 experiments on the streak feature alone — testing notification copy, icon design, XP mechanics, and lesson thresholds. By 2022, most DAUs were maintaining active streaks. But the team recognised the crack: people were no longer logging in to learn — they were logging in so they didn't lose. Meaningless sessions. Hollow engagement. Metrics looked healthy; the product experience was quietly eroding. Duolingo's response was Duolingo Max — a GPT-4 powered tier designed to replace external pressure with genuine learning motivation. That is a product team catching its own metric trap before churn arrived. That kind of analytical self-honesty is precisely what I am training myself to practise.

The feature I most want on Duolingo — live AI conversation practice — sits behind the Max paywall. I am on the free tier. When I hit that wall, I didn't feel frustrated. I felt curious about the conversion architecture: at exactly what point in the learning journey does free-tier value become compelling enough that a user upgrades — rather than quietly exits to a competitor?

Duolingo's Q3 FY2024 SEC filing shows 113.1 million MAUs against 8.6 million paid subscribers — a paid penetration of just 8.5%. That means over 91% of the product's daily engagement runs entirely on free-tier psychology: streak mechanics, peer accountability, gamified XP, and the Streak Freeze I've personally used when life interrupted my learning path. Every one of these micro-decisions carries a simultaneous revenue implication and a churn implication. As a free user living inside that calculation every day, I get product education no classroom can replicate.

I am targeting Business Analyst roles alongside Product and Data Analyst positions — and I want to be precise about why that is not hedging. The BA role sits exactly at the intersection I care about most: user behaviour, data interpretation, business system design, and the gap between what a process produces and what a user actually needs. That gap is precisely where product thinking begins.

The 3 to 5 year direction is product leadership — the ability to sit in a room where a metric is declining, ask the question nobody else asked, and build the case that moves the team. I am not waiting to develop that capacity on the job. I am practising it now — in every product I open, every business headline I interrogate, every case study I dissect. The analytical layer — SQL, Python, Tableau, business acumen — is the foundation. Product leadership is what gets built on top of it, deliberately and without shortcuts. This portfolio is the first exhibit of that proof.