How Startups
Become Defensible

A network effect exists when each additional user makes the product more valuable for all existing users. This is the most powerful defensibility mechanism because it creates a self-reinforcing cycle: value attracts users, users generate value, which attracts more users.

Network effects come in distinct varieties. Direct network effects (each new user directly adds value to others, as in a messaging platform) are the strongest. Indirect network effects (more users on one side improve value for users on the other side, as in a marketplace) compound more slowly but are still structurally significant. Data network effects (more usage generates better algorithmic outputs, which attract more usage) are increasingly the dominant form in software.

A data advantage exists when a company's accumulated data asset enables it to produce outputs — predictions, recommendations, fraud signals, personalisation — that a competitor without the same data history cannot replicate regardless of engineering quality.

Data advantages are often underestimated in the early stage because the advantage is invisible when the data set is small. The compound effect only becomes apparent at scale. This means data defensibility must be designed into the product architecture from day one — not added retroactively when the competitive threat materialises.

The diagnostic question is whether the product produces a proprietary signal — not just data that could in principle be acquired from a third party. Data bought from a vendor is not a moat. Data generated uniquely by your product's usage is.

Switching costs make it expensive — in time, money, risk, or organisational disruption — for a customer to move to a competitor even when the competitor's product is nominally superior. High switching costs change the decision calculus: a customer will tolerate a meaningful product gap rather than absorb the cost of migration.

Switching costs accumulate through data lock-in (historical records stored in a proprietary format), workflow integration (the product is embedded in operational processes), trained behaviour (users have invested time in learning a specific interface), and contractual mechanisms. The strongest switching cost structures combine multiple layers simultaneously.

The design principle is to create genuine value through depth of integration — not artificial friction through bad UX. Customers who stay because migration is genuinely costly are loyal in the economically meaningful sense.

A flywheel is a system in which growth in one component drives growth in another, which feeds back to the first. Unlike a network effect (which operates within a single dimension) a flywheel operates across multiple business components simultaneously — supply, demand, economics, and product quality can each reinforce the others in a closed loop.

The Amazon flywheel is the canonical example: lower prices attract more customers, more customers attract more third-party sellers, more sellers increase selection, better selection attracts more customers, higher volume enables lower costs, which enables lower prices. Each element feeds back to the others. A competitor entering at any single point of the loop faces the full cumulative force of all the others.

Designing a flywheel requires identifying which component will generate the most compounding force per unit of investment — and sequencing early product decisions to activate that component first.