How to Do SEO on Your Website: Six Tiles, Three Moves Each

Why a homepage rewrite moves almost nothing

The homepage is one tile out of six, and on a catalog site it is rarely the page that converts. Buyers arrive from search on a category or product page, judge it in about three seconds, and leave if the title promises something the page does not show. Rewriting the homepage five times still routes that traffic into category pages with duplicate titles and no internal links.

The same mistake appears in reverse on service sites, where every quote request passes through a contact form nobody has titled, and on content sites, where forty articles point at nothing. SEO on a website is a routing problem before it is a writing problem. Fix the tiles the traffic actually touches, then connect them.

One more thing worth saying plainly: you do not need new pages to do this. In our experience most sites need fewer URLs with stronger tiles, not more URLs with the same three missing moves repeated two thousand times.

The six tiles and the job each one does

Every tile has a different job in the buying path, so the moves differ. Treat them as one system: category pages sell the shortlist, product pages close the comparison, articles create the questions that categories answer, and about and contact pages remove the last hesitation.

Three moves per tile, written out

This is the working table we use on calls. Left column is the tile, the next three columns are the moves in the order we apply them, and the last column is the signal that tells you the tile is done.

Notice that only two of the eighteen moves are about writing new text. The rest are structure, links and speed, which is why a tile can be finished in a week by one person with template access.

Which tile goes first

Do not start at the top of the list. Start where two things overlap: pages that already receive organic entrances, and pages with the worst three moves. That intersection is where a week of work produces a visible change, which is what buys you permission to continue.

1. Pull entrances, not rankings. Sort landing pages by organic entrances for the last ninety days. A page with entrances has demand; a page with a good rank and no entrances has a click problem, which is a different ticket.
2. Score the three moves. Give each page a 0 or 1 for query match, internal link and friction. Pages scoring 1 out of 3 with high entrances are your first tile.
3. Fix the template, not the page. If twenty category pages share the same weakness, the fix belongs in the template. Change it once, re-crawl, and the whole tile moves together.
4. Then take the neighbour. After product comes category, because the links you just added point upward. Sequence tiles along the path a buyer walks.

The pre-launch bar: seven checks before a tile ships

Every tile goes through the same bar before it is marked done. It takes about ten minutes per template and it catches the errors that quietly undo a week of work.

• One H1 per page
• Unique title under 60 characters
• Readable URL, no parameters
• One canonical, self-referencing
• Alt text on the main image
• No broken link in the path
• Structured data validates

Run the bar on a staging URL first. If a template fails two checks, fix the template before you touch content, because every page built on it will fail the same way. If it fails only on individual pages, log them and fix them in the same week, since a half-shipped tile is harder to measure than no tile at all.

Four mistakes that make tiles look finished when they are not

• Titles written for every page, links for none. A tile without internal links is a leaf with no branch. It will rank for its own name and nothing else, and the next tile inherits no authority.
• Category copy dumped below the grid. Buyers and crawlers both read top-down. Sixty words above the products do more than six hundred below them.
• Facets left indexable. Colour, size and sort parameters multiply one category into hundreds of near-duplicates. Canonicalise or block them, then re-crawl to confirm.
• Contact page treated as a design asset. Every extra field, captcha and required phone number costs conversions. The contact tile exists to be finished, not admired.

How we work a tile, start to finish

We run the same eleven-day loop per tile: baseline on day one, template changes by day four, content by day seven, re-crawl and check bar by day nine, measurement note by day eleven. You always know which tile is open and what is blocking it.

1. Baseline and tile pick. We export entrances, score the three moves, and agree the first tile. If templates are locked or the catalogue is large, our On-Page SEO Services team handles the crawl, canonical and render work before copy changes.
2. Template and copy in one pass. Titles, H1s, opening blocks and internal links ship together, so the tile is measurable as a unit. Product and category copy comes from our Technical SEO Services desk, written to the query and the spec sheet.
3. Check bar and re-crawl. Seven checks, one crawl, one short list of leftovers. Nothing is marked done on trust.
4. Measure and hand over. We compare the same pages and the same queries, then leave the next tile queued with its own baseline already recorded.

A six-week tile rollout on a fittings store

A furniture fittings store came in with 240 products, 31 category pages and a homepage that had been rewritten twice in a year. Organic entrances existed, but they landed on category pages with identical titles and no links to the products that closed best.

Week one went to the product tile: model numbers into H1s, spec tables converted from images to HTML, breadcrumbs restored. Week two took the category tile, where 31 titles became 31 distinct queries and each page gained six product links plus a short block above the grid. Week three connected the two and removed 180 faceted URLs from the index.

Weeks four to six covered article, home, about and contact. The article tile brought four buying guides that answered installation questions, each linking down to a category. The contact tile dropped a nine-field form to five and put WhatsApp beside the form instead of the footer. Category entrances rose, product pages started appearing for model-number queries, and form completions improved without any change to ad spend.