SEO
How We Run a Technical SEO Audit, and Measure Every Fix
The audit and measurement routine we use on our own sites: what we check first, how we decide what to fix, and why every change gets a 28-day before-and-after.
By TeknoesisPublished Updated 9 min read
Most SEO advice skips the dull part: knowing what changed and whether it helped. Our routine is built around that. We take a baseline, fix things in order of impact, change one thing per page at a time, and check the result 28 days later.
1. Take a baseline before touching anything
Export the last 28 days from Google Search Console: queries, pages, impressions, clicks, click-through rate and average position. Note the date. Also check Google's ranking update history. If a core update is rolling out, or finished within the last week, rankings are moving for reasons that have nothing to do with you, so don't judge changes in that window.
Without a baseline, every later conversation turns into "it feels better". With one, you can say which change moved which page.
2. Can Google crawl and index the site?
- One canonical host. Pick www or the bare domain, redirect the other with a 301, and make every canonical tag, sitemap URL and internal link use the same one. This very site had canonicals pointing at the bare domain while the domain redirected to www; the rebuild fixed it.
- Status codes. Crawl the site and look for 404s from internal links, redirect chains, and pages returning 200 with "not found" content.
- robots.txt and noindex. Check nothing important is blocked or noindexed by accident, and don't block good bots.
- The sitemap. It should list only canonical URLs that return 200, with real last-modified dates.
- URL Inspection. For your key pages, check the Google-selected canonical matches yours.
3. Duplicates and overlap
Look for the same content at several URLs (parameters, trailing slashes, http and https) and for several pages competing for one query. The fix for the first is a canonical or a redirect. The fix for the second is usually to merge the pages into the stronger one and 301 the other, rather than writing yet another page.
4. Speed and Core Web Vitals
Start with field data (the Core Web Vitals report in Search Console, or CrUX in PageSpeed Insights), because that's what real visitors experience. Then use Lighthouse on a production build to find causes. The usual culprits are an oversized hero image, a headline that fades in from invisible, web fonts that shift the layout, and third-party scripts loading too early.
5. Structured data that matches the page
Organization, BreadcrumbList and Article or BlogPosting cover most sites. Mark up only what a visitor can see on the page. Don't add review stars about your own business, and don't add markup for rich results Google no longer shows, such as FAQ and HowTo on ordinary business sites.
6. Improve what already gets impressions
The highest-leverage pages are the ones Google already shows but people rarely click: good impressions, an average position between about 8 and 20, or a low click-through rate. For each one, make the title and the opening paragraph answer the query directly, remove claims you can't support, and add what's missing. Don't repeat the keyword to fill space, and don't create a new page per keyword or city: Google's spam policies call that scaled content abuse.
7. Change one thing, log it, wait 28 days
Every change goes in a log:
- the date and the URL;
- what changed and why;
- the 28-day baseline for that page.
Then leave the page alone for 28 days and compare. Only change the "last updated" date when the content really changed; bumping it for cosmetic edits teaches Google to ignore it. If a change didn't help, undo it and try the next idea.
What we don't do
We don't buy links, trade links, or link our own sites to each other to pass authority. We don't generate pages from keyword lists. And we don't promise rankings, because nobody can. What we can promise is the work, the log and an honest reading of the data.
If you'd like this routine run on your site, see our technical SEO service.