❝Story14:00
How Too Much Scope Burned Him Out at Facebook
Bangaly ran Facebook's people-recommendations team with 30 engineers and huge resources, but the scope was actually too big to deliver without burning out. Layered on top of two or three concurrent manager changes, he had no one to lean on and couldn't see the forest for the trees. There was nothing wrong with the team — the confluence of excessive scope and no stable manager just wasn't a fit.
- Led people-recommendations with ~30 engineers (15 ML, 15 frontend/backend) and lots of resources
- Too much scope became a problem — he felt under-resourced to deliver everything without burning out
- Two or three manager changes happened concurrently, leaving no one to help him navigate
- The environment variables of scope and manager going wrong at once made a great team a bad fit
“I had 30 Engineers I was working with 15 machine earning engineers and 15 fun and”
“in some ways it was too much scope and to me that was problematic”
#career#scope#burnout#facebook
❝Story22:00
The Instagram Colleague Who 'Won Trust in Every Moment'
Early at Instagram, Bangaly watched George Le, who had stood up the growth team, run meetings. George was a masterful listener who could hear a room, recast what was said, and crystallize action items so everyone left with buy-in and clarity. Bangaly studied and emulated that skill — proof that PMs should be students of practice, not just theory.
- George Le stood up Instagram's growth team and left ~six months after Bangaly joined
- His gift was listening — hearing the room and recasting it back so everyone felt good
- He crystallized action items and named people's contributions, winning trust in every moment
- Bangaly consciously stole and emulated the skill — be a student of practice, not just theory
“he had just a really incredible ability to be in a room and hear what was going on to recast it back and make sure…”
#communication#instagram#meetings#learning
❝Story39:30
Instagram's Signup Flow Had Zero Logging in 2016
When Bangaly joined Instagram in January 2016, the signup flow had literally no instrumentation — only counts of how many people started and finished, with eight unknown steps in between. The first move was to add logging fast to find where users dropped, while running quick tests on obviously broken parts of the experience in parallel. A reminder that even iconic products can have shocking measurement gaps.
- In Jan 2016 Instagram's signup flow logged only top-of-funnel and bottom-of-funnel counts
- There were eight steps in the funnel with no visibility into the drops between them
- First priority was instrumenting the funnel as fast as possible to find the leaks
- In parallel they ran tests on obviously broken parts of the experience to capture quick wins
“the signup flow at Instagram had literally no logging to it it had logging of how many people started and how many people ended”
#instagram#growth#onboarding#instrumentation
❝Story01:01:00
Instacart Buried Reorder Behind 7-8 Clicks
Bangaly was shocked that early Instacart made reordering hard, since data showed that after five orders, ~90% of a customer's cart is the same staple items. Reordering took seven or eight clicks to find, and you had to re-add the entire order rather than pick out the milk and blueberries. Fixing the 'adjacent order' — making the next order easy — was likely one of Instacart's biggest growth wins.
- Data showed that after five orders, ~90% of a customer's order is the same
- Reordering was buried behind seven or eight clicks
- You couldn't take pieces — you had to re-add the whole previous order
- Making the next order easy (the 'adjacent order') was a major growth lever
“after five times you go to instacart 90% of your order is the same”
“you had to dig and find it was like seven or eight clicks”
#instacart#growth#retention#product
❝Story01:10:30
Instagram's Untold Growth Engine: SEO and Celebrity Partnerships
Everyone credits Instagram's growth to virality and word of mouth, but Bangaly says celebrity partnerships and SEO were core and rarely discussed. A savvy partnerships team taught celebrities to use the platform, the news linked to their profiles, and launching web created canonical SEO tables plus embeds across the internet. Launching web alone increased Instagram's growth by 10% the minute it shipped.
- Celebrity partnerships were a critical but rarely-discussed growth funnel — they taught celebrities to use Instagram
- Celebrities set the norm for how the platform was used and got picked up by news and media
- Launching web created canonical SEO tables plus embeds across news sites, feeding a whole SEO engine
- Launching web increased Instagram's growth by 10% the minute it launched
- Invitations, celebrities, SEO, paid media, and own content compounded each other at the top of funnel
“another part that goes unspoken still critical to this day was like the celebrity Partnerships was critical”
“the idea of launching web really drove and actually increased Instagram's growth by 10% the minute we launched it”
#instagram#growth#seo#partnerships
❝Story01:17:30
The 'Connections Pivot' That Doubled Instagram Retention
Early Instagram treated every follow as equal, so new users loaded up on celebrities — then made their first post to an echo chamber where none of their friends followed them, and quietly churned. In 2017 the team ran the 'connections pivot,' convincing Kevin and Mikey to prioritize real human connections at signup over celebrity recommendations. Retention doubled over about a year and a half — a bigger driver of Instagram's growth than Stories.
- Early Instagram treated every follow as equal, so new users followed mostly celebrities
- Users posted into an echo chamber — no friends followed them — and churned months later
- The 2017 'connections pivot' prioritized real human connections at signup over celebrities
- Retention doubled over roughly a year and a half — a larger factor than Stories in Instagram's growth
“we had to do this thing we called it the connections pivot”
“we literally our retention doubled over the course of like a year and a half”
#instagram#retention#growth#onboarding
❝Story01:20:00
The Login Fix That Added 15-20M Users a Year
Hundreds of thousands of people a day couldn't get back into their Instagram accounts, costing 10-12 million users a year to 'account access churn.' Rather than stop people from logging out — many share phones or use prepaid data — the team built an omni-box login, trusted-device text codes, and saved credentials. It added an estimated 15-20M monthly actives a year and, unexpectedly, spawned the multiple-accounts feature.
- Hundreds of thousands of people a day couldn't get back into their accounts; 10-12M users lost per year
- They preserved the ability to log out (many share phones or ration prepaid data) and focused on getting people back in
- Fixes: an omni-box for email/phone/handle, trusted-device text verification, and saved credentials/passcode
- The fix drove ~15-20M extra monthly actives a year
- Getting people back in unexpectedly spiked content creation from second/third accounts, birthing the multiple-accounts team
“hundreds of thousands of people a day could not get back into the account”
“we were losing 10 12 million people a year Year from what we called account churn account access churn”
#instagram#growth#retention#login
❝Story01:26:30
'I Have 10 Friends Named Amit Kumar' — Facebook's India Problem
Facebook's people graph looked broken in India: users had far fewer friends in common and lots of friending/unfriending, and no data test cracked it. Bangaly flew engineers to homes in Delhi and Mumbai every few months to watch people make friends, and learned the Western-centric profile (school, job title) was useless there — users ignored it and scanned photos for a friend's car or animal instead. The most common name, Amit Kumar, had ~250,000 real people a month, making recommendations culturally impossible without a rethink.
- In India, friendship connections averaged ~7 friends in common vs ~22 in the US, with heavy friend/unfriend churn
- Data tests failed, so the team did on-the-ground understand work in Delhi and Mumbai homes every three months
- Facebook's Western-centric profile fields (school, job title) were irrelevant; users scanned photos instead
- 'Amit Kumar' was the most common name — ~250,000 real people a month — making name-based recommendations nearly impossible
“this guy's named Amit Kumar I have 10 friends named Amit Kumar right like what is this page going to tell me”
“the most common name was Amit Kumar there was like 250,000 people a month who used AIT Kumar like who were like real people”
#facebook#india#user-research#growth
❝Story01:29:30
Fail Corner: Misreading Instacart's DNA
Bangaly's candid failure was Instacart, where he arrived with a big product vision but was out of tune with the company's operations-heavy DNA. They needed a tactical, hands-on, in-the-weeds leader; he wanted to build systems, people, and processes, and it became loggerheads. His lesson: do understand work before taking a job — and talk not just to current employees but to people who've left, who give the rawer, truer picture.
- Instacart was his weakest stretch — he brought a big vision but misread the ops-first company DNA
- They needed tactical, in-the-weeds leadership; he leaned toward systems, people, and processes
- The mismatch became loggerheads — he felt he wasn't delivering what they wanted or being supported
- Lesson: do understand work before taking a job, and talk to people who've left, not just current staff, to triangulate the truth
“my time at instacart was probably not my best time”
“the people left will tell you what the worst version their or their version”
#career#instacart#failure#job-search