Source: YouTube (20VC with Harry Stebbings), March 2026 Link: https://www.youtube.com/watch?v=6ctOnY1WmBw
The Eight Modes of Defensibility
Gokul's framework for scoring company durability (riff on Hamilton Helmer's 7 Powers). Score each 0-1. Four or more = pretty damn secure. One or less = in trouble.
- Data Mode — Proprietary data nobody else has. (Spotify's decade of listening behavior powering Discover.)
- Workflow Mode — How deeply embedded you are in operations. Weak alone, but depth matters. (NetSuite runs your business = strong. Zendesk = lighter.)
- Regulatory Mode — Licenses, capital requirements, multi-year procurement contracts. (Coinbase: money transmission licenses state-by-state.)
- Distribution Mode — Proprietary/exclusive distribution channels. (Intuit: trained a network of CPAs to only use QuickBooks.)
- Ecosystem Mode — Third parties built on your platform. (Shopify: 100K+ developers, every merchant uses 5-6 third-party apps.)
- Network Mode — Marketplace density, reputation history, liquidity. (DoorDash: can't wipe-code liquidity and density.)
- Physical Infrastructure Mode — Atoms over bits. Hardware, logistics. Harder to displace.
- Scale Mode — Costs so low by virtue of scale that replication is impractical. (Amazon, TSMC.)
Brand is explicitly excluded. Gokul argues brand mode is weakening in B2B because switching costs are approaching zero (data portability + AI clones). Consumer brand may still hold.
Pricing Insights
Seat Pricing Is Not Dead
- Seats provide predictability for enterprise buyers — that's why ChatGPT Enterprise still uses seats.
- But seats break when the product's core value is work done on your behalf, not access.
- Two categories: Access Products (seat-based) vs. Work Products (outcome-based).
- Harvey (legal AI) should price per contract processed, not per user.
Outcome-Based Pricing for AI Products
- If the user isn't the constraint anymore — the work output is — charge for outcomes.
- Workflow-based or transaction-based approaches.
Raise Prices Until Customers Negotiate
- PayPal raised prices 5 times in 3 years because of stickiness.
- If customers never push back, you're leaving money on the table.
Margins Don't Matter Early — Defensibility Does
- Don't obsess over margins in years 1-2 (or even 3-5).
- Two ways to improve margins: (1) increase prices, (2) decrease costs.
- Inference costs dropping = costs will naturally improve.
- The real question: Do you have enough product leverage to raise prices later?
- "Good companies, if they're defensible, will have the ability to increase margin."
Systems of Record Must Commoditize a Complement
- Salesforce/NetSuite must decide: is the profit pool in the data or the workflows?
- If workflows → make data storage free, switch to outcome-based model.
- If data → give away agentic workflows for free, build better products on their data.
- "They can't just wait around for other people to build on top of them."
Multi-Product Strategy ("Compound or Die")
- You cannot be a single-product company.
- Product #2 must emanate naturally from Product #1 (adjacent, not separate).
- Not every product needs to generate profit. Some are profit pool products, some are retention products. Teams must know which is which.
- Square went from 1 product (payments) to 11 products each doing $50M+ revenue.
- Key metric: median number of products used per customer — more products = more retention.
AI-Era Product Strategy
Bolt-On AI That Works vs. Doesn't
- Bolt-on AI has a real ceiling unless you reframe what the product does.
- Adding AI search = upgrade. Rebuilding search as an experience with new UX primitives = transformation.
- You must fine-tune models for your customer base — not just slap GPT/Anthropic on top.
- Re-evaluate every single interaction: "What has changed with new model capabilities?"
- Product roadmaps can't be too long — model capabilities improve every 6 months and blow out your roadmap.
Vertical SaaS in AI Era
- Vertical companies must own the full stack — one function in a vertical is not enough.
- ServiceTitan: 32 products, still sub-$10B. Vertical is harder to scale huge.
- The unlock: vertical software now takes over labor, not just tooling. Target BPO + human capital spend.
BPO Displacement Sequence
- Cut outsourced BPO spend first (easiest — same service, higher quality, 20-30% cheaper).
- Don't replace people who leave (attrition).
- Layoffs (still further out).
Non-Consumption Markets
- The biggest hits create entirely new behaviors that didn't exist before.
- Granola (note-taking), Gamma (presentations), Shopify (anyone can sell) — products so remarkable they create a category from zero.
- "These are also the biggest misses. If the bet doesn't play out, they're screwed."
- Google, Facebook, Uber — all non-consumption markets.
On Market Sizing
- You can't skip it. Do bottoms-up within specific segments.
- Any customer base with 10K+ customers needs segmentation.
- Talk to customers to understand budget and propensity to pay.
- Biggest miss: Shopify — dismissed TAM of "e-commerce merchants" but missed that Shopify enabled anyone to sell.
- Quins (D2C): dismissed the category, missed 35-40% repeat purchase rate that was in the blurb.
On Retention as the North Star
- Durability > growth rate. A company doing triple-triple-double-double with excellent gross + net revenue retention beats a 10x grower with bad retention.
- The Jasper cautionary tale: 1 → 100M → back to 40M. Tire-kickers in prosumer products.
- Two fundamental indicators of business quality: gross retention and net revenue retention.
On Selling / Liquidity
- Fred Wilson's strategy: sell a third, hold a third, trade a third (for liquid assets).
- Focus on go-forward IRR, not just MOIC. A 7x over 20 years is teens IRR — not venture returns.
- If go-forward IRR is below fund target at any liquidity event, sell at least a portion.
- If company would return 20-40% of the fund, you owe it to LPs to sell a piece.
Quotable Lines
- "If they never negotiate, you are leaving money on the table."
- "You cannot be a single product company."
- "The most ambitious entrepreneurs are finally tackling the hardest problems."
- "Brand mode is not so relevant anymore" (in B2B).
- "Durability and defensibility is much more of a worry [than margins]."
- "If you have zero [modes], you're screwed."