Naval - Good Products Are Hard to Vary
Why this is in the vault
Naval applies David Deutsch’s “good explanations are hard to vary” criterion to product design. The frame: a good product is one where you cannot remove or change any component without breaking the whole. Bad products are loose collections of features that could equally well have been arranged differently.
This is a sharp criterion for product decisions because it provides a falsification test. If a feature could be moved, removed, or replaced with no consequence, it is not load-bearing - and a product full of non-load-bearing features is incoherent.
Core argument
Deutsch’s epistemology argues good explanations are reach-extending and “hard to vary” - changing any element breaks the whole. Naval extends this to physical and software products. A great product is a tightly coupled set of decisions where each part justifies the others.
The implication for builders: the test for a feature is not “would someone use this?” but “is this load-bearing for the product’s coherence?” Loose-coupling features dilute the product. Tight-coupling features reinforce it.
Key claims
Under 15 words each.
- “Good products are hard to vary.”
- “If you can remove a feature without consequence, it shouldn’t be there.”
- “Coherence beats completeness.”
- “Loose-coupling additions dilute the product’s signal.”
RDCO mapping
- Sanity Check format coherence. Each issue is a tight argument with a hook, a frame, and an implication. The “Data Dots” feature, the CTA placement, the section structure - each must be load-bearing or it should be cut. The hard-to-vary test is the editorial gate: if I removed this section, would the issue still cohere as the same kind of artifact? If yes, the section was decoration.
- MAC info-product scope. The temptation in info-products is to add features (templates, bonuses, courses) until the bundle looks bigger. Naval’s frame: bigger is not the goal; harder-to-vary is. A MAC pack where every component is load-bearing for the founder transformation is more valuable than a pack with twice as much loosely-related material.
- Bookstore-for-agents wedge definition. When defining the wedge, the test is hard-to-vary. Could we remove the agent-distribution element and still have the same product? Could we remove the bookstore framing? If the answer is yes, the wedge is not yet coherent.
- Vault note discipline. Each vault assessment should be hard-to-vary. The Why-in-vault, Core argument, RDCO mapping, and Related sections are load-bearing; remove any one and the note degrades. Avoid filler sections.
Related
- ~/rdco-vault/06-reference/2026-05-05-jorgenson-almanack-of-naval-ravikant.md - philosophical foundation in Part I judgment chapters
- ~/rdco-vault/06-reference/2026-05-05-naval-find-simplest-thing.md - the iteration-from-simple companion essay