What a design system audit costs, and what you get

Published prices run from $1,000 to $75,000 and up. That spread isn't seniority — it's that the word "audit" is sold against three genuinely different products, and the tables don't say which one they're quoting. I charge $6,000 for a fixed two-to-three-week scope with a written result, which is my rate card and therefore an interested number; the rest of this is how to tell the three apart so you can price whichever one you actually need.

The published prices don't agree with each other

I opened three agency pricing pages on September 7, 2026 and wrote down what each one states. All three sell audits, which is the normal situation — the people who publish prices for a service are the people selling it.

Each ladder is internally sensible. Orbix's four tiers escalate on a real variable: how many people you talk to. Design Studio's mid tier names its inputs precisely — 20 to 40 session recordings, five to eight live testing sessions. Tenet is upfront that it sells the same work on three different pricing models. None of this is disreputable, and I'd take any of these tables over a page that refuses to name a figure at all.

The problem is what happens when you put them side by side. "Enterprise audit" means $18,000 at one and $75,000 at another. Four weeks buys the top tier in one table and the middle of another. A buyer comparing quotes is not comparing prices for a thing — they're comparing three vendors' private definitions of a word.

Three products, one word

Nearly all of the spread resolves once you separate what's actually being sold.

A heuristic review. One experienced person works through your product against a usability framework and writes up what's wrong. It's the cheapest tier in every table above, it's genuinely useful, and it's the one where seniority matters least — a good contractor produces most of the value. If someone quotes you $20,000 for this, they're quoting a different product.

A research audit. Session recordings, moderated sessions with real users, sometimes competitive benchmarking. This is expensive for an honest reason: recruiting, scheduling and paying participants costs real money and real weeks, and no amount of judgment removes that cost. When the tables above cross $8,000, this is usually what crossed it.

A design-system audit. Not the experience — the machinery. How much of the interface actually runs on the system, where it's been escaped, how far the brands have drifted from each other, what's defined but unused. Different question, different skill, different output, and it's the one most often bought by accident when a team says "audit" and means "our UI is inconsistent and I don't know why."

Buying the wrong product at a fair price is the common failure here, not overpaying for the right one.

What my $6,000 buys, and what it doesn't

It's on my rate card: fixed scope, two to three weeks, fixed price, a written result. That's a UX audit, a strategy sprint, or a design-system assessment — one of them, scoped before we start.

Being straight about what that number is not: it does not include recruiting or paying research participants, moderated sessions, or accessibility certification. If what you need is fifteen real users in front of the product, the $8,000–$18,000 tier in those tables is the honest answer and I'd tell you to buy it there. My number is lower because the scope is narrower, not because the same work costs less when I do it.

What it is: one person with fifteen years of this deciding what to fix first, and writing it down in a form your team can act on without me in the room.

What a design-system audit actually measures

Here's the part that's usually hand-waved. A design-system audit produces counts, and the counts are cheap. I ran the check over this site's own system today:

• 465 files scanned, 466 components, 111 screens.
• 8,439 token usages across 138 of the 143 tokens in the contract.
• 5 tokens defined but never used.
• 247 findings, of which exactly 1 renders wrong.

A script produced all four of those on demand, and any competent engineer could write it. If an audit's deliverable is a list like this, you have bought a report that a tool generates, and you should pay tool prices for it.

The deliverable is the next paragraph.

Four of those five unused tokens are the vibrant set. It's an optional namespace — brands inherit it rather than declaring it, and nothing here needs it yet. It isn't dead; it's deliberate, and removing it would be the audit doing damage. The fifth, radius/nested-md, is a genuine orphan: its five siblings are all in use and it never got picked up. That's a real, small finding.

The 247 is the same story. One of them is an actual defect — a raw rgba() shadow in a lightbox component, which sits outside the token contract and won't follow a brand when the theme changes. The other 246 are a review list: arbitrary values that are probably fine and are worth a human glance. An audit that reports "247 issues" is technically accurate and useless. An audit that reports one defect, one orphan, and 246 things you can safely ignore is worth paying for, and it takes judgment rather than a script to tell you which is which.

That's the whole product. The counting is free. The sorting is what you're buying.

Make the deliverable something a team can build against

The clearest test I know of an audit's value: on Monday morning, can an engineer open the deliverable and start work without asking the auditor a question?

A deck of annotated screenshots usually fails that test. It describes problems in the auditor's language, ordered by where they appeared in the product rather than by what to do first, and it needs its author present to be actionable — which is often the point, commercially.

Navigate is the version I'd point at. It was a real-estate outreach tool whose prototype worked but whose core actions had been buried by growth. The engagement was an audit, not a build: I worked the product the way one of their agents would — spinning up area campaigns, targeting homes on a map, tracking responses — found where it broke down, and delivered a prioritized spec. That spec became the team's build roadmap. The deliverable was judgment about what to fix first, in the order they'd do it, not a finished interface.

That's what the word should mean. Not a description of the problem — a sequence of work.

Five questions before you sign

1. Which of the three products is this? If the answer includes user research, ask how many participants and who recruits them, because that's most of the price. If it doesn't, the price should be well under $10,000.
2. What is the deliverable, physically? "A report" is not an answer. Ask for a sample from a previous engagement with the client's name removed. Anyone who does this regularly has one.
3. Is the output ordered by priority or by screen? Ordered by screen means you've been handed the sorting problem back.
4. Which findings are generated and which are judged? Any audit that touches a codebase should be using tooling for the counts. Ask which parts a script produced, and treat a vague answer as an answer.
5. What happens if the audit concludes nothing much is wrong? If there's no version of the engagement that ends with "this is basically fine, spend the money elsewhere," you're buying a document with a predetermined conclusion.

When not to buy one

If you already know what's wrong and you're commissioning an audit to win an internal argument, buy the argument directly — a two-hour conversation and a one-page memo is cheaper and more honest, and everyone will know what it was for anyway.

If the product hasn't shipped, there's nothing to audit. What you want is a design review of the thing you're about to build, which is a different and cheaper conversation.

And if the real problem is that nobody owns design direction, an audit will describe that problem accurately and then leave. That's an ongoing role, not a document, and buying the document first mostly delays the decision by a month.

Related: what fractional design leadership costs, what scales across fifty products, the Navigate audit, and how I work.

Prices above were read from each publisher's own page on September 7, 2026 and will change.