On-page SEO
optimization in New York

A catalogue with filters does not have the number of pages somebody designed. It has the number of combinations the filters allow, and that is a much larger figure. Every colour, size, brand and price band that can be selected together is a distinct address.

Most of those addresses should never be treated as pages. They are views. A view is a way of looking at a list that already exists, and it earns a page of its own only when people are actually looking for that combination.

The test is demand, not tidiness. If people search for a brand within a category, that combination deserves a real page with its own title, its own introduction and a link somebody can find. If nobody searches for four filters stacked together, it should still work perfectly for a visitor and be clearly marked as a variation rather than an original.

Pagination is the second source of the same problem. Page two of a list is a genuine page and needs to stay reachable, because it is how a crawler finds the items on it. It also should not be competing for the same query as page one. The usual mistake is copying the introductory paragraph onto every page of the sequence, which produces twenty near-identical pages by hand.

Sorting is a third. The same list ordered by price is the same list. Ordering is a preference, and a preference does not create content.

Then infinite scrolling, which looks elegant and can hide the entire catalogue. If items only appear when somebody scrolls, and there is no paged set of addresses behind that behaviour, the items further down may not be found at all. The fix is to keep a real paginated structure underneath and let the scrolling sit on top of it.

Internal linking decides most of this quietly. The combinations you link to from menus, category pages and body copy are the ones that get crawled often. The ones reachable only by clicking four filters are seen rarely, whatever the technical markup says. Linking is a stronger signal about importance than anything you can declare.

Empty results need attention too. A filter combination with nothing behind it should say so, offer a way back, and not sit in the index as a thin page waiting to be found.

We audit this by crawling the site the way a machine does and counting what comes back. The number of addresses a catalogue produces is almost always a surprise to the team running it, and the count is the start of the conversation.

Our on-page service for New York sites covers page titles, heading structure, content, internal links and schema, with every change tied to query data. Three tiers are published: on-page fixes ~ $550, full SEO optimization ~ $1,100, large sites with custom setup ~ $2,300. Timing is agreed in the estimate after we see the site.

Search Console holds the most honest description of a page. It lists the queries the page already appears for. Open the Performance report, filter to one page, switch to Queries, and turn on impressions, clicks, CTR and average position. Export it. On-page work on an existing page starts from that file.

Sort by impressions and read the top fifty rows. They fall into groups.

Strong position, weak click-through. The page is found and passed over. That points at the title and the snippet, and at whether the first screen answers the query in plain words.

Positions 8 to 20. Here the page is relevant enough to be considered and loses to pages that cover the subject in more depth. These rows are the cheapest wins in the export. The fix lives in the body copy, on that exact question.

Then the queries that do not belong. A service page drawing impressions for a definition, or for job searches, is telling you its copy reads as something else. Sometimes the right edit is to cut the wording that attracts them.

Device and country filters split the export further. A query that performs on desktop and fails on mobile points at the layout of the first screen. Impressions from countries the business does not serve mean the page ranks for wording it does not need.

Regex filters make a long export manageable. One pattern matching how, what, why and cost separates the informational rows from the commercial ones in a single step.

The report has limits worth knowing before anyone draws conclusions. Rare queries are withheld for privacy, so the query rows never add up to the page total. Average position is averaged across every impression, including those deep in a carousel. History goes back sixteen months and stops.

So we export before touching anything. A baseline saved on the day of the edit keeps the query list as it stood. Wait long enough and that window drops out of the interface.

Every edit also gets a line in a change log: date, URL, what changed. Search Console keeps no such record. Three months on, nobody remembers which week the heading was rewritten.

Comparisons run over equal periods with weekdays matched: 28 days against 28 days. Shorter windows swing with weekends and holidays. An edited page takes days to be recrawled and weeks to settle. The first real read comes about a month after the change.

Seasonality is the other trap. Against the previous period, a December or August window can mislead badly. Year-over-year windows correct for it once a page has a year of history, and the export for both goes into the same sheet.

The row count gets watched too. A page that starts appearing for more distinct queries is being read as covering more of its subject. Clicks come later. Sometimes they never come, when the new queries are informational and the answer sits in the results page itself.

Branded queries go in a separate column. They move with reputation and advertising, and they flatter any report they are mixed into.

Pages with almost no impressions are a separate finding. Either the page is not indexed, or it is indexed and never chosen. The Page indexing report tells the two apart. They need different fixes.