When a design system stops working, the first proposal is usually a new one. It feels decisive. It is also the most expensive way to find out what went wrong with the old one, because a rebuild started without that answer tends to repeat it.
Most systems that are not working have enough components. What they are missing is agreement: on which tokens are real, on who decides what gets added and on why people quietly build around the library instead of with it. None of that is fixed by drawing the components again.
Four questions before any rebuild
An audit is not a review of how the components look. It answers four questions, and each one needs evidence rather than opinion.
1. What exists
Every component, token and pattern, in the design files and in the code, and where the two have drifted apart. The drift matters more than the inventory. A button with four states in Figma and six in the codebase is two buttons, whatever the library says.
2. What gets bypassed, and why
Look at what actually ships and compare it with what the library offers. The gap is usually the most useful number an audit produces. Then ask the people who work around the system why they do it. The reason is rarely carelessness. It is usually a missing state, a component that cannot take the content it is given or a rule that was never written down.
3. Who decides
How does a new component get added today? Who approves it, and does anyone? A system with no answer here will drift again whatever is rebuilt, because the drift came from governance, with components on top.
4. What it would cost to fix
Every gap, prioritised and costed. This is the step that turns frustration into a decision. Some gaps are a day's work and remove most of the bypassing. Some are a quarter's work and remove very little. Without the costs every gap looks equally urgent, and a rebuild looks like the only way to deal with all of them at once.
The answer can be: change nothing
An honest audit has to allow for the recommendation nobody asked for. Sometimes the system is sound and the problem is one missing component and an unclear review step. Sometimes the right move is to stop maintaining a system that one product and two designers never needed. An audit that can only ever recommend a rebuild is a sales pitch.
What it costs to skip it
Extracting a system from years of accumulated screens is slow, and it is slower when the extraction repeats the decisions that caused the drift. A fortnight of looking first is cheap by comparison. That is unglamorous work, and it is the part most rebuilds leave out.
We run this as a fixed-scope, two-week engagement, written up so that whoever does the work afterwards can use it, including your own team. The design system audit page sets out what it covers.