Agent Skills: Product Strategy + Moat Stress Test

Product Strategy + Moat Stress Test

Turn a vague challenge, a goal list or a gut feeling into a written strategy and a verdict on whether your advantage is real. Built from 909 insights from 236 Lenny's Podcast guests. Lead team: Roger Martin, Richard Rumelt, Hamilton Helmer.

When to use

• You were handed a revenue target, an OKR list or 'priorities' and told to produce a strategy.
• Leadership asks 'what is our moat?' and the honest answer is features, speed or the team.
• You must choose where to play: new segment, upmarket, new product, adjacent category, new country.
• A competitor, a bundler, a foundation-model lab or a platform shift just changed your odds.
• Roadmap debates keep turning into arguments about values, which means the strategy is missing or unwritten.
• A team needs a vision (6 months to 5 years) that engineers and designers can act on.
• You are about to pitch or defend a strategy to a CEO, board or investor.

Step 1 — Diagnose (ask before answering)

First read anything the user attached (strategy doc, deck, OKRs, roadmap, pricing page, repo README). Then ask only what you cannot infer, max 4:

1. Stage? Pre-PMF, post-PMF growth, or scaled/incumbent. Why: Helmer says think about power from day one but only as tilted probabilities; Gustaf Alströmer and Uri Levine say pre-PMF the strategy is one priority. It decides plan depth.
2. What triggered this? (a) a target or list with no logic, (b) choose where to play or expand, (c) a competitive or AI threat, (d) a moat claim to defend, (e) a vision gap. Why: it picks the play in Step 2.
3. Is AI your product, your threat, or both? Why: switches on the AI moat tests (Step 3, Play 3).
4. What time and authority do you have? An afternoon, a week, or 8-12 weeks; and who is the one accountable decider. Why: Chandra Janakiraman's process scales from days to weeks; Nickey Skarstad says one leader decides, the team informs.

Step 2 — Pick the play

| If… (situation) | Use | From | Why | |---|---|---|---| | Handed goals, targets or 17 priorities | Play 1: Diagnosis, guiding policy, coherent actions (action agenda) | Richard Rumelt | Goals are ambitions, not strategy; the missing parts are diagnosis and what you will not do | | Need to choose segment, positioning, how to win vs rivals | Play 2: Choice cascade + can't-or-won't test | Roger Martin | Forces linked choices on what you control that compel customer action | | A deck claims a moat, or an AI feature 'can't be copied' | Play 3: Benefit + barrier + materiality test, then AI add-ons | Hamilton Helmer, Anish Acharya, Scott Wu | Most claimed advantages fail one of the three tests | | Need a real strategy for a team/area within weeks | Play 4: Small-S sprint or foundry | Chandra Janakiraman, Rumelt | Procedural; ends in 3 pillars and explicit non-focus | | Tempted by a new product, market, upmarket move or acquisition | Play 5: Right-to-win gates | Jeetu Patel, Dharmesh Shah, Vijay Iyengar | Expansion kills cores; gates catch it early | | Team needs a picture of the future, not a plan | Play 6: Vision in one page | Ebi Atawodi, Melissa Perri, Jenny Wen | Vision must be concrete, lofty and free of today's limits | | Pre-PMF, tiny team | Founding hypothesis sentence only, then execute | Jake Knapp + John Zeratsky, Gustaf Alströmer | A full strategy assumes you can do many things at once | | Under attack from a bundler or platform | Play 3 (barrier + business-model attack check), then Play 1 | Drew Houston, Michael Truell | UI features get copied and bundled free; the economics are the target |

Most requests need Play 1 or 2 plus Play 3. Do not run every play.

Step 3 — Run it

Play 1 — Action agenda (Rumelt)

1. State the problem, not the goal. Ask 'what makes it hard?' until the real obstacle is named; the first answer is usually a symptom. Rewrite 'grow 50%' as the obstacle to growing.
2. Pressure-test the diagnosis against real constraints (headcount, runway, platform dependency). Will Larson: most bad strategies come from willful disbelief, e.g. assuming you can do every project at once.
3. Find the crux: the hardest part you could actually get around. Look for the insight that goes around it; do not pick a strategy from a list of common strategies.
4. Write the guiding policy in one paragraph, including what you will NOT do. Test: could a reasonable person disagree with it (Adam Mosseri)? 'Be the best' fails. Principles need a trade-off: we will sacrifice X for Y (Tomer Cohen).
5. List 3-5 coherent actions for 6 months to 2-3 years, phrased as things we will do, not goals. Conflict check: does any action fight another (Rumelt's example: grow faster while raising return on equity)?
6. Connect the dots (Jackie Bavaro): business target → ~20 things you could do → which fit and why → fill the missing links. Cut to 3 pillars max (Janakiraman).
7. Done when three people from different functions can say what they will stop doing on Monday (John Cutler: crisp pillars often leave people unclear) and you can state what changes for customers (Ami Vora).

Play 2 — Choice cascade (Roger Martin)

Work the five questions iteratively, toggling until they reinforce each other; do not spend long on aspiration first.

1. Winning aspiration: anchored in what customers lack that you can provide. 'The market is big and we want some' is rejected.
2. Where to play: which customers (and which you will not serve), finished product or component, channel, value-chain stage. Segment on needs and willingness to pay, not persona; Madhavan Ramanujam's test: product, sales, marketing and finance must act differently per segment (60% of firms claim segmentation, about 10% have it).
3. How to win: better or cheaper, with a written theory of why customers will perceive it. Coin-flip test: would customers flip a coin between you and the nearest rival? If yes, you have no strategy. Low cost needs dominant scale; niche cost leadership almost never works.
4. Capabilities: what must you do that rivals can't or won't (won't = it cannibalizes them; Martin's fault lines). Interlock capabilities so copying means wholesale change.
5. Management systems that build and maintain them (the moat needs upkeep).
6. Compress to the founding hypothesis (Knapp/Zeratsky): If we solve [problem] for [customer] with [approach], they will choose it over [competitors] because of [differentiator 1] and [differentiator 2]. Add a backup approach. Startups with no scale: list your unique insight, motivation and capabilities first.

Play 3 — Moat stress test (Helmer, with Martin and AI voices)

1. List 3-5 claimed advantages from the deck, board memo or README, plus the things the team spends most time on. Ask of each: castle or shack?
2. Benefit: does it let you charge more or cost less than the best alternative? Quantify.
3. Barrier: why can't a funded rival copy it? Use can't-or-won't. Jason Droege: does 'they can't' still hold in two years? If not, plan for margin compression. Nikita Bier: incumbents typically take 12-24 months to respond, so a barrier shorter than that is a head start, not a moat.
4. Materiality: does the effect move margins or returns materially into the future? Flywheels exist but rarely matter; network effects are not network economies; data scale economies are rare because rivals are usually within shouting distance of your scale.
5. Classify into counter-positioning, scale economies, switching costs, network economies, process power, branding, cornered resource. Availability by phase: take branding and process power off the table at the start, resource only in rights-based businesses; startups almost always begin with counter-positioning (the incumbent can't copy without hurting its own business).
6. Disqualify: operational excellence and speed (necessary, copyable), bought brand recognition, copyable UI (Blockbuster copied Netflix's), team quality, a prototype that took three months and now takes a weekend.
7. Verdict per claim: Castle (benefit + barrier + material) / Head start (barrier under 24 months) / Shack / Unproven. For each non-castle write the one thing that would upgrade it.

AI add-ons (run all five):

• Replicability: none of the classic moats were based on how hard the software is to build (Anish Acharya). Could a rival rebuild it in a week (Cowork took about 1.5 weeks)?
• Proprietary signal: do usage, traces or feedback give you data others lack (Varun Mohan, Peter Deng's data flywheel)? Do not train a model for ego where frontier models win.
• Stickiness: does the product accumulate customer-specific knowledge and get used by teams (Scott Wu's multiplayer)? SaaS lock-in lives in configured business rules (Shaun Clowes).
• Layer risk: is the model the whole product or an input? Benedict Evans: commodity layers lack pricing power at steady state; Kevin Weil: vertical, data-gated, workflow-deep use cases are where labs will not go.
• Scaffolding: list what exists only because models are weak; delete it as models improve (Roman Ugarte) and revisit failed ideas after 1-2 model generations (Dianne Penn). Re-run this test at every major model release. If the product is commodity, distribution and defaults become the moat (Evans); plan your exit from any platform you depend on (Brian Balfour).

Play 4 — Team process (pick by time)

• One afternoon: Play 1 solo, then pressure-test with eng and design. Repeated feature fights are a strategy gap; disagreements are different assumptions, so debate assumptions (Bavaro, Mihika Kapoor).
• 2-4 days (Rumelt's foundry): leader plus 7-8 people put 25-50 problems on a wall, pick the 1-2 that are important and addressable, leave with an action agenda.
• 8-12 weeks, small-S (Janakiraman): prep ~4 weeks (working group of engineering, design, data minimum), 3-5 day strategy sprint, design sprint, 1-2 weeks writing, 2-3 weeks rollout. Sprint day 2: brainstorm problems, cluster into 10-15, rename as opportunities, score on four criteria, top 3 become pillars, the rest are explicit non-focus. Day 3: each person writes a plain-language newspaper headline for two years out and you merge them into the aspiration. Do not exceed 2-3 months.
• Always: one accountable decider; write it down so it can be debugged (Larson); test coherence by asking every team what it is doing and why (Melissa Perri).

Play 5 — Right-to-win gates for expansion

Pass all or kill: (1) customer lack, not TAM (Dylan Field: go strength to strength); (2) permission to play and a route to mass distribution (Jeetu Patel), plus reusable components (Geoff Charles); (3) company-product fit and the right sponsorship (Manik Gupta); (4) suites: if you are top three in a category you over-invested, your value is integration (Dharmesh Shah); (5) list the decisions and reports that must now be sliced by product, the third-order cost; (6) core health: if the core churns to competitors and nothing else is best in class, stop (Vijay Iyengar: invest profits, not people); (7) unit economics modeled now, not at scale (Jiaona Zhang, Tom Conrad); (8) staged funding (Megan Cook's Wonder, Explore, Make, Impact, Scale) inside a 70/20/10 portfolio (Varun Parmar).

Play 6 — Vision in one page

Four tests (Ebi Atawodi): lofty, realistic, free of today's technical limits, grounded in a clear user problem. Write once upon a time / one day / because of that / finally, or the headline of the article announcing it shipped. Match the artifact to horizon: a tweet for 6 months, headline or mock for a year, full vision for 4-5 years. In fast AI markets shorten to a 3-6 month prototype (Jenny Wen) and aim fuzzily at 1+ years (Alexander Embiricos). A vision must be concrete, different from today, and say what you will not do (Perri).

Where the experts disagree

1. Does strategy matter pre-PMF? Helmer, Jag Duggal, Shreyas Doshi: always, because great execution times poor strategy wastes years. Gustaf Alströmer, Mayur Kamat, Tara Seshan: speed from hypothesis to data beats grand plans. → Use the founding hypothesis sentence plus a Play 3 tilt check pre-PMF; full strategy docs post-PMF or when irreversible. Default: one page, not a plan.
2. Can AI products have moats? Helmer, Evan Spiegel, Michael Truell: demand real barriers; software is not a moat, ecosystems and being the best are. Anish Acharya, Roman Ugarte, Scott Wu: moats are discovered, stickiness matters. Dan Shipper, Grant Lee: wrappers can be durable. Marc Andreessen: do not prejudge. → Pre-PMF, ship and capture proprietary signals; post-PMF require at least one castle or a stickiness mechanism. Default: run the stress test and re-run it each model release.
3. Watch competitors or ignore them? Patrick Campbell, Varun Parmar, Drew Houston: competition governs growth; run an intel program. Melanie Perkins, Boris Cherny, Tanguy Crusson: look every ~3 months and re-anchor on customers. → Watch when a bundler or well-funded entrant exists; ignore feature-by-feature reaction. Default: quarterly review, away from daily roadmap decisions.
4. Focus or breadth? Drew Houston, Eoghan McCabe, Jag Duggal, Vijay Iyengar: concentrate, kill loved products. Dharmesh Shah, Kunal Shah: go broad when the customer problem is integration or trust is the asset. Alex Komoroske: plant cheap seeds. → Concentrate when rivals out-resource you in most lines; go broad only with a unifying concept and a healthy core. Default: concentrate.

Deliverable

Produce this and offer to save it as strategy.md.

# Strategy one-pager: <product/area>   Date / Owner / Decider

## 1. Context
Stage, horizon, the decision this serves.

## 2. Diagnosis (the crux)
The challenge in one paragraph. What makes it hard. Constraints we are not pretending away.

## 3. Winning aspiration
What customers lack that we can provide. Success looks like (headline, 18-24 months).

## 4. Where to play / where NOT to play
Customers, product, channel, value-chain stage. Explicit non-focus list with the reason.

## 5. How to win
Better or cheaper, and the theory. Founding hypothesis sentence. Capabilities rivals can't or won't build.

## 6. Moat stress test
| Claim | Benefit | Barrier (months) | Material? | Power type | Verdict | Upgrade move |
|---|---|---|---|---|---|---|

## 7. Guiding policy
The rule that constrains future decisions, including what we will not do.

## 8. Coherent actions (max 3 pillars)
Pillar | Actions (6 mo / 1-2 yr) | Owner | What we stop doing | Leading indicator

## 9. Assumptions and kill criteria
What must be true. Cheapest test for each. The result that makes us change course.

## 10. Re-test triggers
Competitor bundles us, model release, churn in core, quarterly review date.

## Next 3 actions
1. ... (owner, date)
2. ...
3. ...

Grade existing work

Score each 1-5; total /40. Return per-criterion score, total, and the top 3 fixes with the guest behind each.

| Criterion | 1 | 3 | 5 | |---|---|---|---| | Diagnosis | Goal or slogan, no obstacle | Problem named, obstacle vague | Crux named, constraints honest (Rumelt, Larson) | | Choices / trade-offs | Everything is a priority | Some cuts, no stated not-doing | Explicit non-focus with reasons; it hurts (Molly Graham) | | Customer anchor | Market size is the rationale | Segment named by persona | Customer lack and need-based segment (Martin, Ramanujam) | | Defensibility | 'Great team, fast execution' | Benefit named, barrier asserted | Benefit + barrier + materiality tested (Helmer) | | Coherence | Actions conflict or are goals | Mostly consistent | Pairwise conflict check passed; actions reinforce (Rumelt) | | Focus | 5+ pillars | 4 pillars | Max 3 pillars, each with owner and stop-doing (Janakiraman) | | Testability | No assumptions listed | Assumptions, no kill criteria | Hypothesis, cheapest test, kill criteria, re-test triggers (Tara Seshan, Ami Vora) | | Behavior change | Cannot say what changes | Changes for the team only | Teams can say what they stop; customers see a change (Perri, Vora) |

Red flags

• Abstract goals, profit targets or ambitions labeled strategy (Rumelt); a list of 17 priorities means nothing was chosen.
• Mission, vision and values written before any problem is named (Rumelt).
• 'Better' as the strategy: a good version of what exists needs a reason rivals can't copy it quickly (Martin, Tobi Lütke, Christopher Lochhead).
• Market size as the reason to enter (Martin, Dylan Field).
• Counting operational excellence, a bought brand, copyable UI or flywheel talk as power (Helmer, Drew Houston).
• Building from competitive fear: Google Buzz and Google+ (Paul Adams), or copying flows blindly, which fails about 95% of the time (Elena Verna).
• Hedging at an inflection point instead of going all in on the market you can lead (Houston).
• Bolt-on products ending up the sixth-best tool in their category (Iyengar); magical thinking on unit economics (Jiaona Zhang); a broken investment-to-return equation no execution can fix (Tom Conrad).
• Strategy written alone, announced, and never tested for confusion; people hide confusion (Bavaro, Skarstad).
• A strategy with no change in what gets built, staffed or funded (Vora).

Receipts

"But abstract, high-level goals, they're not strategy, they're something else. They're ambitions. And ambitions are not a strategy." — Richard Rumelt, Lenny's Podcast (00:17:05)

"Strategy is an integrated set of choices that compels desired customer action." — Roger Martin, Lenny's Podcast (00:17:48)

"Power requires a benefit and a barrier, you have to have something that you do that gives you a better outcome than your competitors, lower cost or higher price, and then something that makes it impossible for somebody else to mimic that." — Hamilton Helmer, Lenny's Podcast (00:29:30)

"So, strategy should hurt. If your goal setting process is not painful, then you're not prioritizing heavily enough." — Molly Graham, Lenny's Podcast (00:51:03)

"moats are most often discovered, not designed." — Anish Acharya, Lenny's Podcast (00:54:54)

"Great execution multiplied by a poor strategy is a waste of everyone's time." — Jag Duggal, Lenny's Podcast (00:43:35)

Go deeper

• references/frameworks.md — every framework in this skill, how to run it
• references/quotes.md — verified quotes with timestamps
• Related skills:
  • prioritize — turn the 3 pillars into a confidence-scored roadmap and OKRs.
  • ai-product-bets — decide which AI features to build and whether they improve as models improve.
  • position — convert the how-to-win into a positioning canvas and strategic narrative.
  • exec-comms — rewrite the strategy for a CEO or board.