Pruning and de-indexing

What pruning and de-indexing is

Pruning and de-indexing is the removal of pages that compete with your own stronger pages for the same search. Some are deleted and redirected to the page that should have had the visit. Others stay live for the people who need them and are taken out of the search catalogue instead, using a noindex instruction. The point is subtraction: one page per search, and everything else pointing at it.

It is not the same job as the strategy pillar’s pruning pass. That one deletes pages that earn nothing for the business at all: expired events, closed offers, nine years of archive nobody reads. This one deals with pages that may be perfectly decent in themselves and are still doing damage, because four of them answer the same question and none of them wins it. In practice this is mostly redirect work and index hygiene rather than editorial judgment.

When you need it

The clearest sign is a search you should own where your own site appears with a different page each month, none of them the one you would have chosen. That is cannibalisation, and it usually comes from years of writing a new post rather than improving the old one.

The second situation is a site carrying more machine-made addresses than real pages: tag pages, filtered product views, paginated archives, print versions, internal search results. Nobody wants to land on any of them, and they sit between a searcher and your actual page.

The third is the aftermath of a rebuild or a shop migration, where the old addresses and the new ones are both in the catalogue.

You do not need this if your site is small and the problem is simply age. A site of thirty pages cannot cannibalise itself. If the honest description of your archive is that it earns nothing and embarrasses you, the strategy-side pruning pass is the cheaper and more useful piece of work. And if you are not yet sure which pages are the strong ones, start with the SEO content audit and come back.

What gets looked at

• Pages competing for the same search. Every search where more than one of your addresses has appeared, grouped, with the evidence beside each group and a decision about which page is the keeper.
• Machine-made addresses. Tag and category archives with one item in them, date archives, sort and filter variations, paginated series, printer-friendly copies. These are de-indexed rather than deleted, because deleting them usually breaks the site.
• Near-identical templated pages. Service-area or location pages produced by swapping a town name, which read as one page to a machine and as filler to a person.
• Pages that should stay for humans but leave the catalogue. Thank-you pages, policy pages, download confirmations, staff-only instructions. They need to exist; they do not need to be findable.
• Pages other sites link to. These are never deleted outright. They are redirected, so that whatever standing the address has earned goes to the page you kept.

What you get

The competing-pages list. One group per search, each with the pages involved, what each one currently earns, the page that wins, and what happens to the others: delete, merge, redirect or noindex. There is an empty column for your objection, and I would far rather argue about ten pages than have you approve two hundred unread.

The redirect map. Every disappearing address with the page it will point to, or marked as deliberately gone where no sensible destination exists. This file is the whole of the technical instruction; your developer can work straight from it.

The noindex list. The pages that stay live and come out of the catalogue, with the reason recorded for each, plus the change your sitemap needs so it stops advertising them.

The record. A saved copy of anything deleted and a dated note of what was removed. Deletions are reversible for about a year when that record exists and permanent when it does not.

A worked example

An illustration, not a client. An equipment hire company in southern Ontario has 380 addresses. Ninety are real pages. Of the rest, 150 are tag and date archives, 40 are filtered views of the hire catalogue, and the remainder are old posts. Eleven addresses mention scissor lift hire: one proper hire page, three location pages differing only in the town, and seven blog posts written over six years.

The decision is that the hire page wins. Three of the posts hold material worth keeping and are merged into it. The other four are deleted and redirected to it. Two of the location pages have genuine local detail and are rewritten to stop repeating the hire page; the third is deleted. The tag archives, date archives and filtered views are de-indexed in one instruction, not deleted. The site ends with 96 pages in the catalogue instead of 380, and one page answering the search that matters.

How it runs

1. A half-hour call. How old the site is, what has been published, what you sell now, and who in the business is nervous about deleting things. No charge.
2. Access. A full list of addresses, read-only access to your search and analytics data, and whatever your platform can tell me about which page links to which.
3. The pass. Five to ten working days for a few hundred addresses. Every group is judged individually; none of this is decided by a rule applied in bulk.
4. Your veto. You strike out anything you want kept, for any reason, including reasons that have nothing to do with search.
5. Delivery. The redirect map, the noindex list and the record, handed to your developer or run by me as content cleanup.

What it costs

Quoted as a fixed price in CAD after the half-hour call, agreed in writing before anything starts. It depends almost entirely on the number of addresses, because the grouping is done by hand. Where this follows an audit, much of the reading is already done and the price reflects that. The pricing page shows the four ways of working here and the published bands.

What happens next

Three things usually follow. The surviving page gets a content refresh, since it now has to do the work of four. The groups that were merged rather than deleted go to duplicate page consolidation so the combined page reads as one piece rather than three stitched together. And the site’s internal linking is redone to point at the keepers. None of that is assumed. This work is complete on its own: a smaller catalogue, one page per search, and every removed address pointing somewhere sensible.