scaling-fallacy

• SKILL.mdcore · 11,611 bytes
• references/lineage.mddoc · 7,101 bytes

# Scaling Fallacy

> **Definition.** The scaling fallacy is the tendency to assume that a system that works at one scale will also work at a different scale — usually larger, sometimes smaller. The assumption is wrong often enough that it's worth treating as a default suspect rather than a default expectation. Designs that work for 10 users frequently break for 10,000; designs that work for one team frequently fail for an entire company; mockups that look right on a single screen frequently fall apart in production data volumes.

The fallacy is named because it's so common in engineering and design discussions: "We've tested this with 5 users in a session and it works great" → "Therefore it will work at full launch volume." The inference is unwarranted. Scale changes things in two distinct ways: it changes the load on the system (volume of data, volume of users, volume of operations), and it changes the kinds of users and use cases the system encounters (broader audience, more diverse use cases, more edge cases).

Both kinds of change can break a design. The skill is anticipating which assumptions might fail at scale, testing them deliberately, and not assuming that small-scale success predicts large-scale success.

## Two kinds of scaling fallacy

The Lidwell taxonomy distinguishes two kinds, and the distinction is useful.

**Load assumptions.** The system is asked to handle more (or sometimes less) data, traffic, transactions, or operations than it was designed for. Performance degrades, components fail, costs explode, infrastructure breaks.