Audits · 01
Most missing tokens aren't missing
An audit flagged four values as gaps in the token set. All four were already there. The audit had been reading pixels instead of bindings.
I was rebuilding a demo app for a government software company against its Figma design. The first token audit came back with four gaps: a gold accent color, a 72px display headline, and two line heights, 1.4 and 1.5. None of them had a matching token.
The obvious next step was to add four tokens. Before doing that, I checked a second way. Instead of comparing the rendered values to the token list, I read the design's actual bindings: which variable each layer in Figma was connected to.
4flagged as missing
0real gaps
The gold was a brand color token. The pixel audit only saw its hex value. The 72px size was the top step of the type scale that already existed. And the line heights weren't a gap at all: the system sets line heights in pixels on purpose, and the ratios were a choice the demo had made on its own.
Audit the bindings, not the pixels.
Why it matters
A pixel audit can only tell you a value exists. A binding audit tells you whether the designer meant it. Those are different questions, and only the second one tells you whether a token is missing.
If I had added the four tokens, the set would have two names for the same gold and a second headline size that sits next to the first. Do that on every project and a token set doubles in size without getting any better. Worse, the next person, or the next AI tool, now has two right answers to choose from.
What I do now
Every audit starts from the bindings. I pull the variables each layer is connected to, then compare those to the token set. A value with no binding is a real finding. A value with a binding is usually fine, even when the pixels look new.
I also write down the false alarms in the audit report, with what each one turned out to be. It saves the next person from chasing the same gold.