How great teams grow, decide, and hire: lessons from Shishir Mehrotra

Executive overview

Most product teams default to funnel thinking and generic frameworks. The real leverage is in finding your growth loops, naming your decision-making rituals, and diagnosing talent on the dimensions that actually predict performance.

Shishir Mehrotra — co-founder of Coda, former VP at YouTube and Microsoft — shares three interlocking frameworks: growth loops (black loop vs. blue loop), eigenquestions for decision-making, and PSHE for evaluating talent.

The question that answers the most other questions is always worth finding first — in growth strategy, product decisions, and hiring.

Black loop and blue loop

• The black loop is the viral share-create cycle: one user makes a doc, shares it with collaborators, some of them create their own. Every doc product has this.
• The blue loop is publishing to the world — closer to YouTube than Google Docs. A third of Coda users enter through this loop, discovering the product via a published doc, not Coda itself.
• Pricing follows loop design: Coda charges only for "makers," not editors or viewers. The goal was zero friction on the share edge — "no dollar signs in the share dial."
• Platform products lose control of the user conversation at scale; the publisher or sharer owns the onboarding moment, not the company.
• To find your loops: listen to how you pitch candidates. The loop usually crystallises under pressure to explain why the business will grow.

Rituals as culture infrastructure

• Bing Gordon's three tests for a golden ritual: it is named, every employee knows it by their first Friday, and it is templated.
• Canonical examples: Amazon's six-pagers, Google's OKRs, Salesforce's V2MOM, Coda's Dory Pulse.
• Dory: questions are submitted and upvoted before discussion begins, so the group decides what to talk about rather than defaulting to the loudest voice.
• Pulse: everyone writes their view privately before others' responses are revealed, eliminating anchoring and groupthink.
• Coinbase's RAPID added a column to pulse: pre-filling each person's role (approver, informed, decider) redirects feedback to the people who need to act.
• Rituals form two ways: organically (a PM solves a meeting problem and it spreads) or deliberately (designed to drive a specific behavior).
• The Switch framework (Chip and Dan Heath) maps to ritual design: direct the rider (teach it in onboarding), motivate the elephant (give it a memorable name), shape the path (template it so it's easy to follow).
• Naming matters disproportionately — a name lets people form identity around a practice and talk about it without explaining it.

Eigenquestions: the question that answers the most other questions

• An eigenquestion is the question whose answer resolves the most subsequent questions on a list — named after the linear algebra concept.
• YouTube's example: instead of debating whether to link out to abc.com for Modern Family, the team asked: "In a decade, will the online video market value consistency or comprehensiveness?" Answering that made the link-out question, the iPhone app question, and a dozen others straightforward.
• Coda's eigenquestion: "Are we a doc that can behave like an app, or an app that can be built like a doc?" They chose doc-first — and named the company "Coda" (doc backwards) to lock the answer.
• The skill is learnable but must be practised in low-stakes contexts. The teleporter interview question is one vehicle: given a fictional teleportation device, you may ask the scientists only two questions before producing a plan.
• Good eigenquestions eliminate the need to ask about size, speed, or features. The sharp two questions are "Is it safe enough for humans?" and "Is it capex-heavy or opex-heavy?" — those two map the entire product strategy space.
• Children are naturally good at eigenquestions; adults have been trained out of it by years of scope-based ranking.

PSHE: evaluating talent on the right axis

• PSHE stands for Problem → Solution → How → Execution — the four levels at which someone can engage with work.
• Junior roles: given P and S, figure out the H and execute. Senior roles: given P, generate S. Staff-level: surface the right P in the first place.
• The "trough of the solution" is the career phase where scope stops being the discriminator — performance is about which level of PSHE someone operates at, not how big a thing they run.
• The framework applies to every function: a great salesperson doesn't just hit quota (E); she enters any territory and identifies which problems to solve and how.
• P-level thinking and eigenquestion ability are highly correlated: both require identifying what is actually important.

Reference checks and interviewing

• Reference checks rank above interview signals. People who worked alongside someone for years know things a 30-minute scenario cannot surface.
• Run references as early in the process as possible — failing someone post-loop on references is a bad candidate experience.
• Best technique: don't reveal what you value. Ask open-ended questions ("who on your team was best at identifying what problems to work on?") and let the reference reveal the answer.
• Draw contrast by offering four personas without ranking them — give the reference an opening so they're not visibly judging the candidate.
• In interviews, balance home court (Airbnb-specific questions), away court (tell me about your own past work), and neutral court (teleporter). Home-court-heavy interviews test whether candidates already have your answers, not whether they can think.
• The presentation at the start of the interview loop is an "away court" window: what candidates choose to lead with, and how, reveals more than prompted questions.