Slow Down to Grow Faster: Bangaly Kaba on Understand Work, Adjacent Users and Career Choices

Kaba's well-known post on choosing where to work grew out of a personal struggle at Facebook, where he felt he was working harder without seeing more benefit. He wanted an objective way to understand what was going on instead of relying on emotion. The result was a framework in which impact is the thing to optimize, and impact depends on two sets of variables: your environment and your skills.

He explains that compensation and level are reflections of impact rather than goals in themselves. The more impact you drive, the more people trust you to operate independently and the more scope you are given. For a product manager, impact can mean creating extreme clarity about where the opportunities are, then delivering quick wins and repeating them to prove the judgment was sound.

On the environment side Kaba lists six variables: your manager, resources, scope, the team, compensation and company culture. Each year he scores them in quarter-point steps from zero to two, where one is neutral, and asks how likely each is to change. Lenny Rachitsky notes that this makes the full score a number out of twelve, and Kaba says the value is in forcing honesty about what limits your impact.

Kaba describes his own Facebook situation: he led people recommendations with a team of roughly 30 engineers, but the scope was too large for the capacity he had, and two or three manager changes left him without a relationship to lean on. Nothing was wrong with the team, he says; it was the combination of scope and manager at the same moment that made the role a poor fit. He stresses that the variables are not equal, and the manager is the one with the power to fix many of the others.

His advice for a difficult manager is to articulate the challenges dispassionately and tie them to their effect on your work. He also suggests asking what your manager is optimizing for, since individual contributors often chase local maxima and miss that something adjacent to their remit sits on the manager's plate. On the skills side he names communication, influence, strategic thinking and execution, with communication usually the most impactful.

To build skills, Kaba recommends being a voracious reader and watching practitioners, such as asking to sit in a respected leader's meetings. He also argues for a stable of three or four mentors you meet on different Fridays, so a cancellation never leaves you without input. To find them, he says, describe the specific challenge you face to people you trust and ask whom they would recommend, rather than cold-asking someone to be your mentor.

Kaba credits "understand, identify, execute" to an older Facebook framework and says he has carried it to Instagram, Instacart, Twitter and YouTube. He contrasts it with the anti-pattern of identify, justify, execute, in which a team picks an idea, pulls data to support it and then sinks enormous time into it. The painful outcome is the celebratory launch dinner followed by flat metrics the next morning.

The core practice is an intentional line item for understanding on the roadmap, as real as the strategy doc or the prototype. Data science might build funnels, engineering might instrument logging, and product might work out a partnership strategy. Run in parallel with execution, each sprint's learnings feed the next, creating what Kaba calls a velocity multiplier. When he left Instagram, he says, 15 teams were each running 12 to 20 experiments a quarter, and roughly 60 to 70 percent of them were positive and shippable.

The balance shifts as a space becomes better understood: when he started at YouTube his team was about 60% execution and 40% understanding, and now it is closer to 80-85% execution. Kaba answers Rachitsky's worry about endless research by insisting that teams must always ship, picking three or four understand projects per roadmap and keeping a portfolio of low- and medium-effort, high-impact bets. Sometimes understanding is simply a cheap test that teaches fast.

Kaba gives a concrete YouTube example: his team ships paid virtual goods in live experiences, and when he joined, nobody had fully mapped the live ecosystem. The first priority was therefore not iterating the product but building the funnel of how many people watch live, see the experience and actually buy. Engineering, data science and product all had intentional roadmap time for that work.

Both hosts agree the approach sounds counterintuitive only because few teams do it. Kaba frames the trade-off plainly: ship more now and see little impact, or ship fewer, better-understood things so that a year later win rate and velocity are higher. Growth, he says, is a series of short-term wins pointed at a longer-term outcome rather than an overnight success.

A second example comes from Instagram, which he joined in January 2016. The eight-step sign-up flow had logging only for how many people started and finished, so the first quarter began with instrumenting every step, running tests on obviously broken parts, and re-evaluating mid-quarter with the full funnel. Rachitsky observes that simply naming this "understand work" gives meaning to activity that teams would otherwise brush aside in favor of shipping.

When Kaba has to improve a team's culture, he reaches for a managing-complex-change framework he thinks he picked up in business school. It has five components: vision, skills, incentives, resources and an action plan. Missing vision yields confusion, missing incentives yield resistance, and a missing action plan yields false starts. By noticing what the team feels, a leader can identify which bucket to work on.

He usually starts with the action plan because it is the easiest lever to pull, for example improving PRD formats, communication or meeting structures. Vision and skills take longer. To help level a team quickly he keeps a deck of skills and frameworks, such as one on product craft that he wrote after arriving at YouTube and finding that his team lacked the shared mental model Instagram had built.

Kaba's earlier careers in education, as a teacher in inner-city Washington and a dean at a boarding school in Switzerland, inform how he coaches. He says walking into a room of 24 seven-year-olds who owe you nothing demands the same communication, vision and influence that a product manager needs with engineers and designers. He also uses Bloom's Taxonomy, from knowledge up through comprehension, application, analysis, synthesis and evaluation, to see where a person is stuck.

Applying Bloom's pyramid, Kaba says common PM failures are understanding a concept without applying it across multiple scenarios, or knowing what to do without synthesizing why it matters to the business. Managers, he argues, have to live at the top of the pyramid for every area they own, synthesizing and evaluating across teams to make bigger-picture connections.

He does not fully buy the idea that a product manager is the CEO, preferring to think of product as a team sport in which the leader is a coach. Not everyone needs to be a star, he says, but you need strong role players who feel valued and sit in the right seat. He also jokes that persuading 24 seven-year-olds was harder than explaining a growth strategy to Instagram's founders.

A basketball-coach friend introduced him to the idea of a coaching tree, which Kaba applies to product leadership. He takes pride in former teammates who now run growth at Stripe or lead other organizations, and he sees their success as his own. Coaching people up, he says, builds a repository of skills and team velocity, and it also frees him to work on bigger problems.

Kaba says growth is largely about understanding the value proposition at every point of the experience, especially in multi-sided marketplaces like YouTube, where creators and viewers have different goals. He pushes teams to dogfood products in the state of an adjacent user: a long-time power user's experience is shaped by years of history, and a freshly created account may see something significantly worse. YouTube's search, for instance, leans on watch and search history that a new user lacks.

He says flywheels also tell teams what to work on and what not to, and reveal what they still do not know about creators, viewers and monetization. At Instacart he asked how groceries fit into daily life: people shop to make Taco Tuesday, not to browse ingredients, so the product should help them assemble a meal. Rachitsky summarizes the lesson as thinking about the value of each interaction and the job to be done in a person's day.

Another Instacart finding was that reordering was surprisingly hard: after five orders, about 90% of a basket was the same, yet repeating it took seven or eight clicks and could not be done piecemeal. Kaba concludes that teams often build the pieces of a flow without verifying that the product delivers the real output. Making the next order easy was, in his view, central to retention.

The adjacent-user idea came out of Instagram's hypergrowth. Kaba says the app went from about 440 million monthly actives in January 2016 to 636 million by year end, and the people using it changed dramatically within months. A telling signal is declining cohort curves: new sign-ups do worse even though the product has not changed, because later users are less tech savvy and arrive from new markets such as India or the Philippines. A registration flow that converted extremely well could drop 15% three months later for that reason.

He insists you must know who uses the product today and who the next user is, then watch, talk to and visit them. At Instacart the original user might have been an office admin, the next a parent of several kids, and later a single person in New York, each with different needs. Kaba calls the concept mandatory for hypergrowth and also useful for mature companies seeking more share of wallet, citing a hypothetical cosmetics firm that lacks certain skin-tone shades.

Where does growth usually hide? Kaba says typically between onboarding and habit formation: many teams do not know their first aha moment, and a leaky bucket wastes top-of-funnel effort. He then lays out Instagram's layered engines: invitations, a partnerships team that taught celebrities to use the platform, web launch and SEO, embeds on news sites, paid media and original content, each magnifying the others. He says launching web lifted growth by about 10% and that the idea came from a colleague before founders Kevin Systrom and Mike Krieger were convinced.

Kaba tells a story that rarely gets airtime: early Instagram treated every follow as equal, so new users were recommended many celebrities. Growth marketing head Rob Andrews noticed retention dipping again after seven to nine months. The team's hypothesis was that people posted for the first time and none of their friends were there, so they felt they were posting into an echo chamber and left.

The fix, which he calls the connections pivot around 2017, was to convince the founders to prioritize real human-to-human connections early rather than celebrities. Kaba says retention improved dramatically, and while Stories brought many people, he argues that this shift in activation paid off enormously. Rachitsky adds that TikTok's strategy of recruiting lower-tier creators worked differently because celebrities were already established on Instagram.

His most impactful launch, he says, addressed hundreds of thousands of people a day who could not get back into their accounts and were lost 28 days later, roughly 10 to 12 million people a year. Because people in places like South and Southeast Asia legitimately log out to save data or share phones, the team kept logging out but unified login into one omnibox and sent a text prompt on trusted devices. They also offered to save credentials at logout. The work recovered an estimated 15 to 20 million monthly actives a year and revealed second- and third-account creation, which led to the multiple-accounts team.

Kaba joined Facebook in 2014 and was responsible for people recommendations globally, with a focus on South and Southeast Asia. In the US a new friendship might share about 22 mutual friends, he says, while in India the figure was around seven, accompanied by lots of friending and unfriending. Tests produced no breakthrough, so he proposed understand work: every three months a team of engineers visited people's homes in Delhi and Mumbai to watch them make friends.

They saw people skip the profile page because its Western-centric fields, such as school and job title, were irrelevant, and people instead looked at photos. Names were another issue: when the team went back to Menlo Park, they found that the top ten most common names were Indian and that Amit Kumar had roughly 250,000 real people using it a month. Kaba says that meant a search for one friend might have thousands of plausible candidates, which forced creative solutions.

In the show's "fail corner," Kaba says his time at Instacart was probably not his best. He arrived with a vision of building systems, people and processes to institutionalize work, but he says the company, strong at operations, needed someone tactical who would get deep into funnel data. He felt neither that he was delivering what they wanted nor that he was supported as he wanted.

His lesson is to do understand work before accepting a job, talking not only to current employees, who tend to share the best version of the company, but also to people who left, who give a rawer view. You then triangulate between both. In the lightning round he adds that in his first 90 days he sits in team meetings, listens, and learns everyone on the team by name and story, which makes people more willing to hear his views later.

Other lightning-round answers include recommending *Range*, *Deep Work* and *Start at the End*, loving the flight-tracking app Flighty, and an interview question that asks senior candidates to stack-rank five job-critical skills from strongest to weakest. He also repeats a line that teams fall to the level of their systems. He closes by saying his time as a boarding-school dean taught him that building products for the world requires being inclusive of many voices.