Your data model is your destiny - Matt Brown's Notes

• Slack, Toast, Notion, Figma, Rippling, and Klaviyo won by choosing non-obvious data models (persistent channels, menu items, blocks, canvas, employee records, order data) that created compounding advantages competitors couldn't replicate without rebuilding from scratch
• With AI commoditizing code execution, your data model becomes your moat—it creates organizational lock-in through workflows, integrations, and institutional memory that can't be easily copied
• Horizontal tools win through technical/interface innovation (Figma's multiplayer canvas); vertical tools win by elevating the right domain objects with the right attributes (Toast's menu-item architecture)
• Audit your data model by checking which database table has the most foreign keys—is that the atomic unit customers actually think in? Test if new features inherit context automatically or require building from scratch
• The paradox: this critical choice happens when you know the least about your market, but that's why it's so powerful—competitors who've already built on different foundations can't simply copy your insight

The author argues that a startup's data model—which parts of reality it emphasizes and how it structures core concepts—is the dark matter holding product-market fit together. While most founders inherit their data model by copying incumbents, the biggest breakout companies trace their success to early, non-obvious data model choices that seemed minor but proved decisive. Slack made persistent channels (not ephemeral messages) the atomic unit, creating organizational memory. Toast built around menu items with embedded restaurant logic (prep times, kitchen routing), not generic SKUs. Rippling treats the employee record as the lynchpin connecting all systems, making each new product module automatically more powerful than standalone alternatives.

This matters more than ever because AI is commoditizing code execution, making technical implementation table stakes rather than a competitive advantage. Meanwhile, vertical markets are so crowded that companies must expand into adjacent products, payments, and even compete with customers' labor. When code is cheap and competition is fierce, your data model becomes your moat—it creates compound advantages through the organizational reality customers build around your architecture: workflows, integrations, institutional muscle memory that competitors can't replicate without starting over.

The path to a differentiated data model depends on your market. Horizontal tools serving broad use cases win through technical and interface innovation (Figma's multiplayer web canvas vs Photoshop's local files). Vertical tools serving specific industries win by elevating the right domain objects (Klaviyo promoting order data to equal status with email metrics). Start by identifying model mismatches in incumbent products—where are customers using spreadsheets and workarounds to make the product match how they actually think? Ask "what's the atomic unit of work in this domain?" not "what data do we need to store?" Then audit whether new features automatically inherit context or require building from scratch. If the latter, you have a product suite, not a compounding data model.