Skip to content

Buyer guide

How to choose a smart contract auditor

Most audit decisions are made on a quote and a logo. The eight checks below are what actually predict whether the review finds the bug that would have drained you.

Written from the reviewer side, which is why it also tells you what to expect from me

Start with what you are buying

A smart contract audit is a short, senior code review with a written deliverable. It is not a guarantee, not a certification and not insurance. Anyone selling it as those things is selling you something else, and the moment an exploit happens the difference matters more than the price did.

What you are actually buying is attention from one person for a fixed number of days, plus a document your team can act on and your investors can read. Everything worth comparing follows from that.

The eight checks

  1. 01

    Who actually reads the code

    Ask who is on the engagement and whether that person is the one writing the findings. If the sales call and the review are done by different people, price the difference in. A firm can be excellent and still send your scope to whoever is free that week.

  2. 02

    The commit under review

    A report must name the commit or the tag it applies to. Without one you cannot tell whether the audit covered the code you are about to deploy, or a snapshot from three weeks earlier that has since been refactored.

  3. 03

    Proof, not adjectives

    Every High and Critical should come with a reproduction: a test, a transaction trace, or exact steps against the commit. A finding you cannot reproduce is a hint, and hints should not be paid for at Critical prices.

  4. 04

    A coverage statement

    The report should say what was not reviewed, as plainly as what was. Contracts out of scope, assumptions about deployment, and areas that needed more time. A report that claims complete coverage of everything is describing marketing, not a review.

  5. 05

    Fix verification

    The cheapest part of the job is finding a bug. Ask what happens after you patch: is each finding re-run against the new commit and marked fixed, partially fixed or open, by the person who wrote it? Without that, the retest is a formality.

  6. 06

    Severity that matches the deployment

    A reentrancy in a function that can only be called by a role you control is not the same finding as one any address can reach. Ask how severity was justified and whether the attack path names a principal who can actually trigger it.

  7. 07

    Someone whose name is on public work

    Disclosed reports, advisories, a CVE, a public writeup with the vulnerable code shown. This is the only reliable evidence of methodology, because it is the part nobody can write after the fact.

  8. 08

    A report your developers can act on

    Ask for a redacted sample. Executable conclusions, a file and line, a concrete fix, and a severity you agree with. If reading it requires a call with the author, it is not finished.

Red flags in a quote

None of these is proof that a reviewer is bad. Each one is a question you now have to ask, and a quote that answers none of them is a price, not a proposal.

  • A quote with no scope document, only a total.
  • No named reviewer, no commit, and no way to tell which code was read.
  • Findings with no reproduction and no file references.
  • A severity matrix that explains nothing about how severity was decided.
  • Guaranteed outcomes. Nobody can promise zero Criticals, and anyone who does is selling insurance they do not underwrite.
  • A discount that appears only after you mention a competitor.
  • A retest quoted as a separate line item, priced by the hour.

The five things to compare before you sign

  1. The commit and the contracts covered.
  2. The number of contracts and lines in scope.
  3. Whether High and Critical findings come with a reproduction.
  4. Whether fixes are verified, and by whom.
  5. Who performs the review, by name.

What a smart contract audit price should buy covers the number side; this page covers the work behind it. Two quotes are only comparable when all five match. That is also why a fixed-fee proposal with a written scope is easier to judge than an hourly estimate: the deliverable is defined before the meter starts.

What to ask for before you commit

  • A redacted report from a previous engagement, or a public writeup with real code in it.
  • The scope document, including what is explicitly out of scope.
  • The start date and the delivery date, in writing.
  • What happens if a Critical is found on day five of a ten-day engagement.

The answers tell you more than the reference calls, because they are checkable now rather than later.

How this site answers its own checklist

The report template is the structure I deliver, published in full, so you can compare it against any other quote. The pre-audit checklist is what I look at and what I expect to be ready. The engagement terms include the fixed fee, the timeline and one fix-review round. And the public work is here and here, including a writeup with the vulnerable pattern shown.

Buyer questions

How do I compare two audit quotes?

Line them up on five things: the commit each one covers, the number of contracts and lines in scope, whether High and Critical findings include a proof of concept, whether fixes are verified, and who performs the review. Two numbers are only comparable when those five match.

Is a big firm safer than an independent auditor?

Not automatically. A large firm buys process, insurance and a brand you can put in an investor deck. An independent buys attention and accountability from one person. What matters either way is who reads your code and whether their work is checkable.

What should an audit report contain?

A scope section naming the commit and the contracts, a summary a non-engineer can read, findings rated by severity with a reproduction and a concrete fix each, and a coverage statement listing what was not reviewed. The report template on this site is the structure I deliver.

How long should an audit take?

Five to ten business days for a small protocol, two to four weeks for a larger one with integrations. A same-week audit of a 5,000-line codebase is a code review, not an audit.

What happens if the auditor finds nothing?

You should still receive the report and the coverage statement, and you should be able to see what was tested. A clean result from a named reviewer is a useful artifact for your users and your investors. An empty page is not.

Do I need an audit before mainnet?

If the protocol holds other people money, yes, and your users, your exchange listing and your investors will all ask for it. What they will read is the report, so the report is the deliverable to judge, not the logo on the cover.

Compare me against the checklist

andrei@dodu.se · written scope, one business day

Get an offer