Thinking beyond frameworks: how to build judgment as a product leader

Executive overview

Most product managers have been trained to follow frameworks as if they were rules, not tools. This produces PMs who can't make decisions without data, research, or a process someone else designed.

The antidote is judgment: knowing when to ship to learn, when to delegate, and when to trust founder instinct versus team expertise.

Frameworks are a toolkit, not a coloring book — the job is to add value to customers, not follow a process.

The zero interest rate phenomenon PM

• Zero interest rates let every startup operate like a well-funded public company with abundant engineering, design, and research support.
• PMs conditioned in this environment default to research and process instead of making decisions under uncertainty.
• Red flag in interviews: "I can't come up with solutions until I see all the data and talk to customers."
• The real PM job is figuring out how to add customer value — not executing someone else's process.
• Advice for PMs who recognise themselves: go to companies where you ship fast and learn quickly; reorient your north star from "following the framework" to "delivering value."
• User research is a scarce resource — reserve it for problems with extreme uncertainty and high leverage. For everything else, ship to learn.

When to use research versus shipping

• Consumer products at scale: bias toward shipping and measuring; bring in research only when data is confusing.
• Enterprise products: customers are sophisticated and rational — talk to them and sales directly.
• Mid-market (e.g. Eventbrite): most nuanced; focus research on the B2B side where problems are big but not yet well-defined, or where the solution is unclear.
• Best researchers proactively tell you when research isn't needed.

How to interview PMs effectively

• Skip work history and rehearsed answers — give real scenarios from the role and watch how candidates approach them.
• Look for: creative problem-solving without data, quick path to signal, holistic thinking about metrics.
• Red flags: solutions requiring months of engineering before any signal; inability to reason without research; no sense of time cost.
• The goal is to see candidates do the job, not perform about having done jobs.

Founder intuition versus team expertise

• Founders who find product-market fit build up hard-to-articulate intuition about customers and product — don't outsource this too early.
• New hires should be directed until they demonstrably make better decisions than the founder would.
• Employees should proactively signal both directions: "I need direction here" or "I've got this, trust me."
• As companies scale, founder expertise broadens and shallows; team expertise narrows and deepens — the balance naturally shifts.
• The failure mode: founder bottlenecks every decision, slowing market launches, restaurant sign-ups, growth channels (the acquired Grubhub competitor example).
• If a founder won't delegate even after you've built the case, find the feedback channel they do respond to — data, peer CEOs, key customers — and build that case instead.

The CPO role

• CPOs are executives first, product executives second — visibly caring about sales, legal, marketing, and finance is not optional.
• Diagnose team and product strengths and weaknesses without bias; lay it out clearly for peers who don't know what "great product" looks like.
• Make your roadmap to improvement legible: "Our OKRs aren't quantitative yet because X — here's the timeline to fix that."
• Leadership mistakes surface fast at the exec level; there's no PIP, only loss of confidence.

Network effects: the three types

• Direct network effects: every additional user improves the product for all existing users (e.g. WhatsApp).
• Cross-side network effects: two distinct user types, each side's growth makes the other side more valuable (e.g. Grubhub: more restaurants attract more diners, and vice versa).
• Data network effects: product quality improves as more data is collected (e.g. Pinterest: saved content improves recommendations for all users).
• Social networks must evolve beyond direct network effects — selling ads creates cross-side effects; personalisation requires data network effects.
• Network effects are the best form of defensibility, but not immunity to disruption.

How DoorDash disrupted Grubhub

• Grubhub was an asset-light marketplace: restaurants used their own drivers; Grubhub connected and processed payments.
• DoorDash and Postmates built managed marketplaces with their own delivery networks — negative-margin businesses subsidised by VC.
• They started in low-density suburbs with restaurants that had never done delivery, growing without competition.
• They also delivered from restaurants without agreements, which dramatically expanded selection when they moved into cities.
• More restaurant selection means more attractive for diners — cross-side network effects reversed against Grubhub.
• Grubhub, now public, had promised high-margin, asset-light growth; competing meant destroying that narrative and building an operations capability it didn't have.
• The only viable move in hindsight: acquire DoorDash early, à la Netflix betting on streaming.
• Key lesson: during existential threats, assume the disruptor is playing an optimal game and react accordingly — underreaction is the default mistake.

Marketplace-to-SaaS versus SaaS-to-marketplace

• Marketplaces adding SaaS (e.g. Faire): use the SaaS tool as a supply acquisition or retention mechanism; works well because you already have the hard part — the marketplace.
• SaaS adding a marketplace (e.g. Eventbrite, OpenTable): works only if the SaaS product already touches the customer's customer, and those customers have a reason to transact with multiple suppliers.
• If end users don't know they're using your tool (e.g. Stripe), there's no path to marketplace.
• SaaS-to-marketplace transitions are possible but slow to de-risk; marketplace-to-SaaS is more replicable and opens adjacencies faster.
• A marketplace drives demand to supply — that's the distinguishing value prop. If supply signs up for tooling or payments alone, it's a SaaS-like network, not a marketplace.

Why consumer subscription is structurally hard

• B2B SaaS has two structural advantages: predictable, rational customers, and net dollar retention (customers spend more over time).
• Consumer subscription has neither: consumers churn faster and don't expand spend.
• To survive, annual retention must exceed 60-70% — a very short list of companies has achieved this at scale (Netflix, Amazon Prime, Spotify, Duolingo).
• Those companies do it through massive content spend or network effects — neither is replicable cheaply.
• Paid acquisition compounds the problem: best customers convert first; every subsequent cohort converts and retains worse until it's unprofitable.
• Default path (freemium + paid acquisition + hope for retention) leads to a predictable death — you can model exactly when.
• The pivot: build organic growth loops, add network effects, or find a supply side that has an incentive to refer users.
• Stay small and scrappy long enough to find what works (Calm vs. Headspace is the canonical cautionary contrast).

AI and the PM role

• If your PM job was filling in frameworks and collecting FAANG promotions, AI will replace it.
• The real PM job — trading off complexity, making decisions under uncertainty, building deep subject matter expertise — is among the least replaceable.
• Current LLMs are trained to sound smart, not to be correct; treat AI-generated analysis with skepticism.
• High-value current use: tedious no-code and low-code tasks (formulas, integrations, formatting) where errors are cheap to catch.