Content Audit

Publishing has a defined start and no defined end. A team that commits to two posts a week usually keeps that commitment, but nobody decides when a post written two years ago gets looked at again. The count goes up and the list never comes down.

A pile built that way loses money quietly. Last year's pricing explainer still names a plan that no longer exists, and it still sits on the first screen of search results, so the person who arrives reads a wrong number and either files a support ticket or leaves. Three posts written separately on almost the same topic eat each other's rankings. When all three compete for one query, the search engine can't tell which is the main one either, and all three end up somewhere in the middle.

The larger problem is that the team can't see the state it's in. When results flatten, the usual response is to write more, because the publishing count is visible and the condition of the pile is not. A content audit is the procedure for opening that pile up once. The idea behind it is to give maintaining what exists the same weight as making something new.

Start with a complete inventory, attach numbers to it, sort by those numbers into four buckets, then handle each bucket. Five steps, usually.

1. Build the list — collect the URL, title, publication date, and target query for every post in one table. Every post, not a sample. The list people carry in their heads holds the successful pieces and the recent ones, which leaves out exactly the old posts that need attention.
2. Attach numbers — record six months of traffic, search position, and conversions for each piece. Short windows make the judgment wobble: look at one month and anything seasonal reads as dead.
3. Sort into four buckets — keep, fix, merge, delete. Traffic plus accurate content means keep. Traffic with stale content means fix. Several posts split across one topic means merge them into one. No traffic and no future role means delete. Search intent, the kind of result someone typing a query actually wants, belongs in this step. If the searcher wanted a comparison table and your post is a concept explainer, the low ranking is about category rather than quality.
4. Handle each bucket — fixing means correcting stale numbers and screenshots and reattaching a call to action (CTA). Merging means folding the other pieces into the strongest one and pointing the merged URLs at it with a redirect; deleting them outright throws away the inbound links and visits they had collected. When two nearly identical pages both have to stay, a canonical tag names the primary one so the search engine counts only that side.
5. Book the next round — put the date for the next pass on the calendar the day this one ends, usually six to twelve months out. Skip this step and the same pile is waiting two years later.

Step 3 takes the most time. Sorting purely by the numbers deletes pieces that get little traffic but that customers read right before signing, so it helps to write down what each piece is there for before assigning it a bucket.

It fits a team that has been publishing for more than two years, has passed a hundred posts, and has watched traffic go flat. Reworking a post already sitting on the second screen of results raises traffic faster than writing a new one, because a piece with a track record doesn't have to build from zero. Search engine optimization (SEO) is the work of tuning a site to surface better in search results, and the parts of it that take the least effort for the quickest visible result are concentrated here.

The reverse case is a team six months into publishing with twenty posts, where it's too early. The numbers to judge with don't exist yet, and twenty is a number one person can hold in their head. Revisiting the content strategy serves better at that stage: without a plan naming who the pieces are for, which topics, and in what order, an audit produces no standard to fix anything against.

The limits are clear enough. This only judges posts that have accumulated numbers, so anything published three months ago is out of scope. Measuring evergreen content, which keeps getting read as time passes, on the same scale as an event recap that runs for three days drops all the recaps into the delete bucket, and some of them were written to run for three days on purpose. Doing it once also doesn't hold. An audit with no second pass a year later is just a record of one year's cleanup.

"Let's run an audit this quarter instead of publishing. Traffic has been flat three months running."

Proposing to spend a quarter on the backlog rather than on new pieces.

"I'll merge these three into one and redirect the other two."

Collecting scattered posts on a topic into a single piece while keeping the old URLs alive.

"Forty posts have zero traffic, but sales still sends twelve of them to prospects."

Pointing out that the list contains pieces the numbers alone would wrongly condemn.

Ravi ran the first one at Pagedeck three years into the blog. The pile came to 210 posts, and once the table existed it turned out that nine of them produced close to half the traffic of the last six months. Of the rest, 88 had single-digit visits over the same period.

Ravi didn't delete all 88. He sent the list to Sofia in sales and asked her to mark anything she links to during conversations, and twelve came back. Those twelve stayed despite zero traffic, because they're written for someone already in a conversation rather than for someone arriving from search. Of the remaining 76, 31 overlapped enough to be merged into nine, and 45 were deleted with their URLs pointed at a related piece.

The most common reading is that an audit is a deletion project. The visible result is a shorter list, so it reads that way. But of the four buckets, the ones that actually raise traffic are fix and merge. Updating the stale parts of a post that already holds a ranking produces a result faster than a new post does, while deleting is closer to housekeeping that makes the next round lighter. A team that understands it only as deletion stalls at the first post nobody wants to throw away, and the whole project stops there.

The other mistake is treating a low number as low value. Traffic counts people arriving from search, so anything never aimed at search scores low by construction — a post laying out the product's philosophy, or an explainer handed over as a link during a sales call. Draw a threshold without that distinction and the pieces sales and recruiting were using disappear together, and you find out who was relying on those links only after they're gone.