Chandra Narayanan on Growth Metrics, Analytics and Building Teams

Chandra Narayanan says a team should focus on meaningful impact rather than activity alone. The Sundial co-founder joins Harry Stebbings to discuss lessons from PayPal, Facebook and Sequoia, including growth metrics, data analysis and hiring.

At PayPal, Narayanan recalls wanting to quit after conflict between senior leaders made his work difficult. His manager Rohan encouraged him to address the problems before deciding. He says he stayed another six to nine months, improved the situation and then left for Facebook with his manager’s help.

Stebbings questions whether fixing a difficult situation is always worth the time. Narayanan describes his experience as character-building and says it later helped him face challenges at Facebook. His personal experience is not a rule that everyone must remain in a harmful workplace.

He recalls presenting data at Facebook that some leaders did not want to hear and says colleagues supported him through the conflict. His broader lesson was to prioritize impact: moving a metric, influencing a product decision or improving a process, rather than merely completing many activities.

He distinguishes motion from progress by asking what changed because of the work. Hard work matters in his view, but repeated activity can plateau; learning and prioritizing the right problem can yield more impact per unit of time.

At Sequoia, his data team worked on sourcing, diligence and helping portfolio companies. He says prioritization mattered: a short list of strong leads could be more useful than hundreds, and diligence could identify reasons to decline a deal before a firm committed time and money.

He recalls finding one company whose marketing had already reached much of its addressable market and another whose newer cohorts retained less well than older ones. Stebbings notes that saying no can miss overlooked winners; Narayanan argues avoiding poor investments also protects investor time. These are examples from his experience, not universal rejection rules.

Narayanan credits Sequoia’s investors, access, diversity of skills and collaborative decisions. He describes circulating material before meetings so participants arrive prepared for a deeper discussion. This positive assessment comes from a former employee.

At Facebook he learned to build high-expectation teams that also cared about people. He praises Alex Schultz and Javier for focusing on impact per person instead of simply increasing headcount. The small-team examples are his recollection, not a controlled comparison.

He illustrates impact per person using market capitalization divided by an assumed engineering headcount, explicitly using rounded numbers for the exercise. The calculation is a leadership heuristic, not a causal measure of what an individual engineer created.

He defines growth as finding ways to scale product-market fit sustainably. A north-star metric should connect to the company’s mission and be something the team can influence; he cites Facebook’s active-user goals and advertiser growth as examples.

A metric may need to change as the product or market changes. He recalls a shift in thinking from monthly to daily active users when mobile made daily use more common. Teams may also discover that their first chosen metric was simply wrong.

Intuition and historical playbooks can be biased, he says, but data also depends on assumptions that may change. He favors making a decision and iterating where possible, rather than assuming either intuition or data is infallible.

A useful hypothesis connects a change in a metric to a possible story: seasonality, a product change, marketing, competition or customer behavior. Narayanan says analysts must move between data and hypotheses, test alternatives and remain aware that the story may still be wrong.

He recalls a PayPal fraud hypothesis that multiple accounts using the same cookie could be suspicious. Investigation also found legitimate families and internet-cafe users sharing computers. Testing both positive and negative cases helped refine the idea rather than assuming every match was fraud.

Not every hypothesis can be validated with existing product data. To understand why advertisers leave, for example, surveys and user research may be needed. He also recalls a rise in time spent that the team eventually associated with unusually cold weather; this is an observed explanation, not a controlled causal study.

He generally advises building a growth function after product-market fit, because growth should scale value people already want. For a consumer product, repeat use without unlimited marketing can be a signal, but he says there is no single definition that fits every business.

He separates product-market fit, scaling that fit, achieving sound unit economics and scaling those economics. A free product can attract users yet struggle once it charges; success at one stage does not guarantee success at the next.

For an initial growth effort he thinks one person may be insufficient: design, marketing, analytics, product and engineering need to work together. If a company can hire only one person first, he suggests a senior leader capable of assembling the necessary team.

He favors a small centralized growth team early so people can share practices while tackling a few company-wide opportunities. As the company grows, members can be embedded in product teams and eventually decentralized. Stebbings questions the budget required; Narayanan says a narrow initial scope can keep the team small.

Hiring ahead depends on stage. He recalls Facebook analytics evolving from counting and dashboards to experimentation and influencing product strategy; hiring only for the last stage too early would have left immediate infrastructure needs unmet. Faster-scaling companies may need to hire farther ahead.

Influence requires both a sound conclusion and an approach the audience can hear. Narayanan says resistance can come from unreceptive leaders, weak evidence, poor communication or bias. He recalls adjusting how he presented data to different colleagues.

He describes founders with different relationships to data: some are knowledgeable and curious, others open to learning, and some resistant or overconfident. He adapts the amount of data and storytelling to the person, while recognizing that influence is difficult and not always possible.

His hiring framework treats a person as a combination of skills, looking for exceptional strengths without major liabilities elsewhere. He wants colleagues who enjoy their work, like and learn from one another and see company progress. Stebbings observes that compensation can also keep people in jobs; Narayanan clarifies that he is describing the team environment he seeks.

For analytics, he looks for the ability to break a business problem into technical and data questions, analyze the evidence, synthesize insights and influence decisions. He reduces the core analysis to two questions: what differs from a useful benchmark, and “so what?”

He illustrates benchmarking with an account of low payment conversion in Germany compared with nearby countries, which he attributes to a missing bank-payment option. The automated captions give inconsistent company names and he is unsure of the reported improvement, so the example should be read as an anecdote rather than a verified causal estimate.

The “so what” question asks whether a change is material to the company’s goal: an opportunity affecting 100,000 users may warrant different priority from one affecting 200. The numbers are illustrative, and materiality depends on context.

For senior candidates he especially values simplification: can they break a complex problem into clear components? He asks candidates to read and synthesize a piece on sustainable growth, then discusses their reasoning rather than treating a single interview answer as definitive.

A hiring mistake he identifies is overvaluing a candidate’s present level or prestigious profile while overlooking the speed at which they learn. He calls these the plateau and the slope. His preference for growth potential is a judgment, not a guarantee of future performance.

He speculates that long-established habits can make change harder with age, while emphasizing that a growth mindset can remain valuable later. This is his personal view and not evidence that age determines adaptability.

He describes his own shifts from oceanography and computing through risk, analytics, venture and founding a startup. Building a remote team across the US and India has been a new challenge, including the difficulty of modeling behavior across distance.

Moving from Facebook’s abundant data to earlier companies required simpler frameworks. He describes breaking a revenue model into users, time spent and advertising, and using business-type formulas to identify which data matters. He says this is a way to reason with limited data, not a substitute for evidence.

When someone struggles, Narayanan looks for a skill gap, knowledge gap or values and culture mismatch. He says a role or team change can sometimes help, and managers should diagnose the cause instead of assuming every performance problem has the same solution.

He thinks performance improvement plans often start too late, after a manager has already given up. Earlier, direct feedback and support may give someone a more meaningful chance to improve. He also acknowledges that a hiring process can miss important gaps and that interviews remain uncertain.

He groups executives by the situations in which they excel: turning a troubled organization around, improving an adequate one, or taking a good one further. He argues that matching the person to the company’s needs matters more than a prestigious résumé; his categories are a heuristic, not fixed limits on individuals.

He recalls arriving at Facebook with strong opinions from PayPal and receiving feedback that he needed to listen and bring colleagues along. His lesson is that past experience can create unwarranted certainty about a new organization.

In quick fire, he warns founders not to hire purely for a famous profile. He describes using meditation and recounts a childhood experience with hypnosis and stuttering; this is a personal story, not evidence of a treatment effect or medical advice.

He would like growth teams to use counter-metrics: PayPal revenue growth and fraud prevention, for example, can pull in different directions, so teams should watch both outcomes. A single improving metric can hide a harmful trade-off.

Narayanan says growth is harder than he initially thought because talent, infrastructure, prioritization, leadership support and execution must work together. He closes by praising Mark Zuckerberg’s ability to translate a high-level direction into a detailed strategy and team roadmap; this is his former colleague’s assessment.