The keep, kill, edit, combine content audit

By Matija Konjić

Every site that has published for a few years carries the same inventory problem: hundreds of pages, a handful doing the work, and nobody quite sure which is which. New content keeps shipping because shipping feels like progress, while the archive quietly drags, splitting authority, cannibalizing queries and teaching search engines that most of what the domain publishes goes nowhere.

The content audit is the fix, and the version we run is called keep, kill, edit, combine, the same method inside our content marketing process. It assigns every page one of four verdicts, with a reason, and turns the archive from a liability back into raw material. This guide is the full method: the data to pull, the verdict logic, the execution order, and the mistakes that turn pruning into self-harm.

Why audits pay before new content does

On an established site, the fastest wins are usually already published. An aging post that once ranked returns to form faster than a new page can climb, because it holds history, links and indexed trust that a fresh URL starts without. Editing twenty of those beats writing twenty new pieces on speed alone, and the audit is how you find the twenty.

The negative side pays too. Thin, overlapping and dead pages are not neutral: they dilute internal links, compete with your own better pages for the same queries, and sit in exactly the pattern search systems now demote at scale. Removing them concentrates what the domain already earned, which is why audited sites so often rise before a single new piece ships.

A shape we see repeatedly makes it concrete. A blog holding two hundred posts finds sixty genuinely working, eighty worth editing or merging, and sixty doing nothing but splitting the same queries three ways. Six months after the audit ships, the site holds one hundred forty pages and more organic traffic than the two hundred ever produced, with the biggest gains on pages the audit merely edited. Nothing about the pattern is exotic; it is what concentration does.

When to run one

Yearly is the rhythm for most sites, with two extra triggers: after a core update that moved the site meaningfully in either direction, and before any big content investment, because a roadmap built on an unaudited archive plans in the dark. The first audit is the heavy one; every following pass is maintenance on a known inventory.

Scope it honestly by size. A fifty-page site audits in a day. A few hundred pages is a week of focused work spread across a fortnight. Thousands of pages means auditing by section, money pages and their clusters first, the long-tail archive after, because a partial audit finished beats a total audit abandoned at row four hundred.

The data that decides

Verdicts come from evidence, and four exports supply nearly all of it:

• Search performance. Twelve months of clicks and impressions per page from Search Console, which separates working pages from decorative ones.
• Rankings. Which queries each page holds, and at what position, to spot the almost-ranking pages an edit can push over.
• Links. Internal links in and out, plus external links earned, because a linked page is an asset even when its traffic is not.
• Business fit. Whether the topic still maps to what you sell, which no tool exports and only you can answer.

Assemble the four into one sheet, one row per page, and the audit becomes reading rows rather than debating opinions. A few hours of assembly for a mid-size blog, most of it export and paste.

Reading the sheet fast

Sort the sheet three ways before judging anything. By clicks descending, and the keep column mostly writes itself. By position, and the eleven-to-twenty band surfaces the edit queue with the fastest payback. By topic, and the combine clusters become visible as three or four rows circling one query with none of them winning it. An hour of sorting answers more than a week of page-by-page deliberation, and it keeps the verdicts anchored to columns rather than nostalgia.

The four verdicts

Every row gets exactly one verdict, and the definitions keep the meeting short.

Keep is for pages that rank, convert or earn links: they get left alone beyond fresh internal links pointed at them. Edit is for right topics wearing wrong execution: stale facts, thin sections, missed intent, all fixable on the existing URL. Combine is for the clusters of thin overlapping posts a younger site wrote around one topic: they merge into a single strong page, with redirects carrying their history. Kill is for pages with no demand, no rankings, no links and no business job: they get removed, with redirects to the closest relevant survivor.

The judgment calls

Edges exist and two rules settle most of them. A page with links and no traffic leans keep or combine rather than kill, because earned authority is expensive to abandon, exactly the logic behind valuing links. A page with traffic and no business fit leans edit toward relevance rather than kill, because demand you already own is cheaper to redirect than to rebuy. When a row genuinely splits the room, mark it edit and revisit next pass; the audit does not need heroic certainty to work.

Seasonal pages get one extra courtesy: judge them on their season rather than on a flat twelve months. A guide that earns its keep every November looks dead in a June export, and killing it in July is how sites donate their December traffic to competitors. Mark the seasonal rows, compare season against season, and verdict them on the window they were built for.

Executing without breaking things

Order the work by risk. Kills and combines ship first, because they are mechanical: remove, merge, redirect one-to-one, and update the internal links that pointed at the removed pages. Edits run as a production queue afterward, prioritized by how close each page sits to winning: positions eleven through twenty first, stale top-tens second, everything else behind them.

Redirects deserve pedantry. Every removed URL points to the closest surviving relevant page, never in bulk to the homepage, and the map gets kept, because a redirect chain discovered two years later costs an afternoon a proper map would have saved. Internal links get rewired the same week, so authority keeps flowing through the new structure rather than dead-ending into redirects, the same discipline internal linking as internal linking always demands.

Announcing nothing

Resist the urge to relaunch loudly. Pruned and merged pages need no announcement, and edited pages earn their attention back through the rankings rather than a campaign. The one audience worth notifying is your own team, so nobody links to a removed URL next month or resurrects a killed topic in next quarter’s roadmap out of habit.

What an edit actually includes

A real refresh reworks rather than repaints: facts and examples updated, thin sections rebuilt, intent re-checked against what now ranks, internal links added both ways, and the piece re-run through the same editorial bar new content passes. Changing the date without changing the substance is the industry’s least convincing trick, and readers and rankings treat it accordingly.

Capacity math for the edit queue

Size the queue against real production capacity, exactly as a roadmap of new pieces would be. A proper refresh costs half to two-thirds of a new article’s effort, so a team shipping four new pieces a month can clear roughly six edits in the same hours. Alternate months of drafting and refreshing, or reserve a fixed slot per week, and the queue drains within two quarters instead of becoming a second backlog that shames everyone quietly.

Measuring the audit

Baseline before touching anything: organic clicks to the whole archive, rankings on the terms the audit targets, and index counts. Then read the results on the honest clock, six to twelve weeks for the pruning effects, a quarter for the big edits, and expect the signature pattern: total pages down, total traffic flat then rising, average performance per page sharply up.

Keep one number for stakeholders: traffic per published page. It is the metric audits move most reliably, it survives seasonality better than raw totals, and it tells the real story, the same site doing more with less, which is what concentration means.

Annotate the timeline while you measure. Note the week kills shipped, the week each major edit relaunched, and any core update that landed mid-window, because attribution arguments die when the dates sit on the chart. The audit that cannot show its work gets credited with nothing, and the one that can usually earns next year’s budget on the spot.

The mistakes that turn pruning into harm

Audits fail in famous ways, all avoidable. Mass-deleting by a traffic threshold alone kills linked pages and future winners along with the filler; the four-export sheet exists precisely to stop one column deciding alone. Skipping redirects leaks years of accumulated authority through 404s. Auditing without a content strategy behind it prunes the garden and plants nothing, which looks tidy and compounds nothing. And treating the audit as a one-time cleanse rather than a rhythm guarantees the same meeting in two years, with a bigger sheet.

One more failure hides in good intentions: auditing while publishing continues unexamined. If the pipeline keeps shipping pieces that would fail the same audit, the archive refills with future kill rows at exactly the pace the cleanup empties them. Fix the bar at the front, the briefs, the editorial pass, the strategy filter, and the audit becomes a yearly trim instead of a recurring rescue.

Run it properly, once a year, on evidence, with redirects mapped and edits queued into real production, and the archive becomes what it should have been all along: the cheapest source of growth the site owns. If you would rather hand the sheet, the verdicts and the follow-through to people who do this weekly, that is exactly what our content strategy engagements start with, because a roadmap built on an audited archive is the only kind that starts from the truth.