How to Do SEO for Your Website: Record, Change, Retest
Doing SEO on your own website is a bookkeeping problem before it is a tactics problem. Write down what your site looks like today, change one thing, then come back on a date you set in advance and compare. Everything else, from plugins to checklists, only works if that record exists.
Printable. Three pages. No tool subscription required.
How do you do SEO on your own website without guessing?
Export the current numbers for the pages you plan to touch, write the change and the expected effect in a dated row, apply one change, then compare the same URLs on the date you already booked. A written baseline is what separates a test from a hope.
Answer blockBuild a baseline you can defend four weeks later
A baseline is not a screenshot of your traffic graph. It is a file, dated, with one row per URL and the same four columns every time: impressions, clicks, average position and the number of enquiries or orders attributed to that page. Anything less and you will not be able to tell which page changed when the total moves.
Export it before you edit a single title. Search Console gives you impressions, clicks and position per page, and your analytics gives you conversions per landing page; join them by URL and keep the file. Ten minutes now saves a quarter of arguing with yourself later about whether the change worked.
Choose the page set before you start. Ten to thirty URLs is right for a solo owner, selected by impressions, not by personal affection. The pages Google already shows are the ones where a small edit produces a measurable difference, and the pages you love are usually the ones nobody searches for.
One variable, one date, one decision
Self-audits fail for a predictable reason: the owner changes the title, the meta description, the images and the internal links in one weekend, then credits the whole lift to whichever change they were most proud of. That is not a test, it is a story.
Change one class of thing across a fixed set of pages. Titles only. Or internal links only. Keep the dates identical, keep the page set identical, and write down what you expect before you look at the result. Predicting first is uncomfortable and it is the only way to learn anything.
Then decide in writing: keep, revert, or extend for another four weeks. A decision recorded is worth more than a ranking gained, because it can be repeated on the next twenty pages.
Four columns, five rows, one page you can print
The worksheet below is deliberately narrow. Current state forces you to write what is true today, target forces a number instead of a wish, action forces one sentence you can actually complete, and retest date forces the follow-up into the calendar. If a row cannot be filled in that shape, the task is not ready to start.
Your notes, this cycle
Print at A4 with default margins. The blank lines are the part people skip, and the part that makes the sheet useful.
Three ways we misjudge our own websites
When you own the site, you also own the story about the site. These three habits account for most of the wrong conclusions we see in self-run audits, and each one has a cheap correction inside the worksheet.
Remembering it as fixed
The title felt wrong last year, you are fairly sure you changed it, and you never checked. Six months later the template reverted during a theme update and nobody noticed.
The row in the sheet is the memory. If the row is empty, the work is unproven, no matter how certain you feel.
Changing everything at once
A free weekend, a new theme setting, five plugins and a copy pass. Traffic moves, and you cannot say which change did it, so you cannot repeat the win.
One variable per cycle feels slow. It is the only way the second page is cheaper than the first.
Comparing to your best month
The baseline everyone remembers is the record month, not the average one. Any normal month then looks like decline and triggers another round of panic edits.
Compare to the dated export, on the same weekday, against pages you deliberately did not touch.
Send us your filled sheet. We will mark it.
We check the baseline, the page set and the retest dates, then tell you which row is measured badly before you spend another cycle on it.
Read your own record before you read another guide
By the time two cycles are logged, the site tells you where to work next. If impressions are flat and clicks are falling, the problem is in the snippets on pages you already rank for. If impressions are rising and conversions are flat, the pages attract the wrong query or the offer is unclear. If everything is flat and the page count is growing, you are publishing pages that never earn a click.
That reading is more useful than any generic priority list, because it is about your pages rather than the average site. It also tells you when to stop: a row that has been retested twice with no movement should be marked closed, not retried a third time.
For a structured second opinion, run the items in our SEO Audit Checklist against the same URL set, then add whatever you find as new rows with their own retest dates rather than mixing them into the current cycle.
Most self-audits stall on internal links
Titles take an afternoon. Content takes a week. Internal links look trivial and get postponed indefinitely, which is unfortunate, because they are the one change a solo owner can make at scale without writing anything new. Every page you add needs an inbound link from a page that already has authority, or it starts life as an orphan.
A practical rule: when you update a page, add three links from related pages in the same section, and make sure the hub links to its children. Six weeks of that habit fixes most crawl and discovery problems on a small site, and it makes the next content launch visible within days instead of months.
When the link work is larger than an afternoon, it is worth treating as its own project rather than a row on the sheet. That is what Internal Linking Services covers: an audit of orphans and anchor text, a link map for the next quarter, and the edits implemented in the templates so new pages inherit them.
How to keep the record honest
Four habits keep a self-audit from turning into a diary of opinions. None of them requires a paid tool, and all of them survive a busy month.
- Date every file in the name, for example baseline-2025-03-01.csv, so a later comparison cannot drift onto the wrong export.
- Write the expected result before the change goes live, in the same row. Expectations written afterwards always match the outcome.
- Keep a control set of pages you never touch in that cycle. When the control moves too, the cause was seasonal, not you.
- Close rows explicitly. Open rows accumulate into a list that feels like progress and measures nothing.
Do this for two quarters and the sheet becomes the most valuable SEO document you own, because it contains what worked on your site, in your market, at your level of authority. That is worth more than any list of best practices, including this one.
Questions from site owners running their own audit
How long should I wait before retesting?
Twenty-eight days for on-page changes, up to fifty-six for new pages. A retest at day seven usually measures how fast Google recrawled, not whether the edit helped.
Do I need a paid tool to record a baseline?
No. Search Console exports impressions, clicks and position per page, and your analytics exports conversions per landing page. A spreadsheet and a naming convention cover it.
How many pages should one cycle cover?
Ten to thirty for a single owner. More than that and the retest becomes an afternoon of sorting instead of a five minute comparison, which is how cycles stop happening.
What if traffic drops right after I change something?
Wait for the retest date before reacting, unless the page is no longer indexed or returns an error. Short term wobble is normal and panic edits destroy the comparison.
Should I test on new pages or existing ones?
Existing pages first, always. They already have impressions and a ranking to improve, so a measurable difference arrives faster than it does from a page starting from zero.
What do I do with a row that never moves?
Close it. After two clean cycles with no change, the hypothesis was wrong. Record why, keep the row as a reference, and spend the next cycle somewhere else.
When should I stop auditing and hire help?
When the bottleneck is implementation rather than decisions, or when a single change needs engineering time you do not have. That is usually internal links, migrations or templated on-page work.