SEO audit: findings ranked by what each one costs

Most audits are a crawler's export: three hundred rows of warnings, sorted by how serious the tool thinks they are. The owner of the site reads one screen and closes it. The example here goes the other way, with five findings in order of the visits each one costs, written for someone who has never opened a crawler. Its figures are estimates, and it says how they were made.

Who it's for

SEO consultants and small agencies, writing to a prospect or to a client in the first weeks of working together. Your reader owns or runs the business. They can approve two days of a developer's time, and they can't judge a canonical tag.

Whoever does the fixing still needs the full crawl. Send it as a spreadsheet alongside, as the sample does with its list of 312 dead addresses. And an audit is a snapshot of one week. If the client wants figures that refresh, that belongs in their analytics tool.

What a good one contains

• The prize as the headline. "Five fixes worth about 2,300 more visits a month" gives the owner a reason to keep reading. The sample turns it into money too: about $5,150 a month.
• A ranking by cost. Biggest loss first, with a bar for each. An owner who stops after the second finding has still read the two that matter most.
• Evidence a non-specialist can check. "In January those 312 pages brought 940 visits. In May they brought none."
• Fix, effort and what to expect, side by side. One developer day against roughly 700 visits back in six to eight weeks. That trio lets someone say yes to finding 1 and "later" to finding 5.
• One line for the developer. The technical version, with its 404s and rel=canonical, sits in a small block under each finding, so the page can be forwarded without a translator.
• What's already fine. Four things nobody should bill for.
• An order of work. Who does what in which week, what it costs ($2,800 of developer time) and the date you'll measure again.

Keep the estimates honest

Every figure after the word "about" is a forecast, so show the working. Near its end, the sample says what was crawled and on which days, explains where each visit estimate comes from, and admits that rankings move. "About 700" is believable, and "712.4" claims a precision no forecast has.

Then cap the list. If the crawl found thirty problems, your job is to pick the handful worth fixing and leave the rest in the spreadsheet. Title each finding in the owner's words, too. "Your workshop pages never say Madison" gets fixed. "Missing local keyword targeting" gets filed.