Choosing a spacing scale for a design system

WHAT THIS TESTS The interviewer wants to see that you treat spacing as a constrained system, not ad hoc pixels, and that you can articulate the tradeoffs between a numeric grid and semantic naming.

A GOOD ANSWER COVERS Start from why a scale exists: limiting choices enforces visual rhythm and consistency across teams. A numeric base-unit grid, typically a 4-point or 8-point system, defines spacing as multiples of a base unit, so values compose predictably, align to a grid, and are easy to reason about mathematically; the downside is that consumers handle raw numbers and very large or very small steps need care. T-shirt sizes expose semantic tokens like sm, md, and lg that read clearly in code and communicate intent rather than magnitude, decoupling design decisions from concrete pixels; the downside is ambiguity about exact values and awkward extension when you need something between md and lg or beyond xl. The strongest approach usually combines both: define a numeric base scale as the source of truth, then layer semantic named tokens that map onto specific steps, giving readable intent backed by predictable math.

COMMON WRONG ANSWERS Allowing arbitrary pixel values with no scale, which destroys consistency. Choosing t-shirt names without defining their numeric values, leaving ambiguity. Picking a base unit so small that everything is just a number with no rhythm.

LIKELY FOLLOW-UPS How do you handle a one-off spacing that does not fit the scale? How does the scale interact with responsive breakpoints and density modes? Should the scale be linear or use a multiplier?

ONE CONCRETE EXAMPLE Define a 4-point base scale of 4, 8, 12, 16, 24, 32, then expose semantic tokens space-sm equal to 8, space-md equal to 16, and space-lg equal to 24. Designers and engineers reference the named tokens for intent, while the underlying multiples keep every layout aligned to the same grid.