The total score is a weighted average across ten dimensions, scaled to 0–100. A common surprise is that a famous company like OpenAI or Walmart scores in the 30–50 range: not because they aren’t valuable, but because their moat is something other than a pure network effect (capital, brand, data, distribution). Look at the radar shape and the Strongest Dimensions list to see what is actually driving the score. If a single dimension feels too low, raise its Priority dropdown to “High” and see how the total moves.

Sarnoff (V = n) describes one-to-many broadcast networks like TV and radio: each new viewer adds a constant amount of value. Metcalfe (V = n²) describes peer-to-peer networks like phones, email, or Bitcoin: every user can connect to every other, so value scales with the square. Reed (V = 2ⁿ) describes group-forming networks like Discord or Slack, where users can form arbitrary subgroups and the number of possible groups grows exponentially. These are theoretical ceilings, not market caps: real networks usually land somewhere between Sarnoff and Metcalfe.

The pink shaded area between the linear baseline and the scored-company curve is the extra value created by the network effect: what this business produces above and beyond what a comparable linear business would produce at the same user count. A wider surplus area means a stronger moat. Note the units are relative, not dollars; the chart is about curve shape, not absolute valuation.

The Priority dropdown lets you weight dimensions you think matter more for your judgment of this specific business. For a crypto protocol, you might raise Protocol and Developer Ecosystem to “High” and drop Local Density to “Low.” For a rideshare app, you’d do the opposite. The default of Normal across all ten is a fine starting point: don’t over-engineer the weights. The shape of the radar matters more than the exact total.

Dominance is not the same thing as network effects. Walmart dominates retail through scale, supply chain, and real estate: not because each new shopper makes the store more useful for everyone else. Google dominates search through data and capital, with weaker direct user-to-user effects than Facebook. The score is measuring the network part of the moat, not the moat itself. A business can be a fantastic investment without being a network effect business.

The presets are opinionated estimates built from the published characteristics of each network: they are starting points for discussion, not authoritative ratings. Bitcoin’s protocol score is high because it is a shared monetary standard; its density score is near zero because Bitcoin doesn’t care where users live. OpenAI’s data and developer scores are high; its marginal cost score is low because GPU inference is genuinely expensive. Tweak any preset after loading it to reflect your own view.

Quite a bit. It doesn’t model anti-network effects (congestion, spam, harassment, declining quality at scale: the “Eternal September” problem). It doesn’t account for regulatory or geopolitical risk, which can break a network overnight. It doesn’t weigh how easily the network can be displaced by an adjacent platform. It doesn’t separate domestic from global effects. Use the score as a structured prompt for analysis, not as a decision rule. A high score is necessary, not sufficient, evidence of a real moat.