Blog Post Refresh and Update Plan
Five documents and four sheets that decompose a traffic decline into its five signatures, then send the rewrite only to the pages a rewrite recovers.
Free download · No account needed
Update Log, cycle 1
What the lost traffic was actually made of
61 of 214 pages declining over twelve weeks, measured against the same twelve weeks a year earlier. A fictional certification publisher, Kelbrook Learning.
| Signature | Clicks lost | Share | Where it goes |
|---|---|---|---|
| Position decay | 12,900 | 37% | Rewrite. The only one that is |
| Presentation decay | 7,600 | 22% | Title and markup. An afternoon |
| Demand decay | 6,130 | 17% | Nothing recovers it |
| Coverage loss | 5,450 | 16% | One canonical, no writing |
| Substitution | 2,780 | 8% | Consolidate, do not rewrite |
| No signature | 40 | 0% | Inside noise |
A third of the loss is a writing job. The other two thirds route to three people who were never going to be asked, and one of those routes is a decision to do nothing and write down why.
The single biggest loser, 6,290 clicks on one guide, gained impressions and held position 3.1 to 3.4. Every click it lost went above the click. Ranked by traffic lost, it is the first page a writer opens.
Kelbrook Learning went into a refresh quarter with 61 declining posts and 34,900 lost clicks, ranked by traffic lost. The top row was a guide down 6,290 clicks, more than any other page on the site. Its impressions had gone up 2.1 per cent and its average position had moved from 3.1 to 3.4. Nothing about that page had got worse. The results page around it had changed, its click-through rate had fallen from 7.9 to 2.6 per cent, and two days of rewriting would have touched a body that nobody had stopped reaching.
That is arithmetic rather than bad luck. Search Console defines click-through rate as clicks divided by impressions, so a click figure is impressions multiplied by click-through rate by construction, and those two factors answer to completely different causes. Rank two hundred posts by clicks lost and the sort has already discarded the diagnosis. Decompose it instead, add average position as the third reading, and five signatures separate cleanly. One of them is a rewrite. The other four are a title, a canonical, a redirect and a decision.
So this pack builds the register with the decomposition in it, assigns a signature to every declining page, and routes. It handles the two traps that make a naive sort actively wrong, then measures the result against a control set of untouched pages so the number survives being questioned. It installs into the marketing workspace, beside the content brief pack, the editorial calendar pack that schedules what the queue produces, the content inventory and pruning pack and the rest of the template library.
What's in the pack
Decay Register
One row per declining page carrying clicks, impressions, click-through rate and average position for both windows side by side, before it carries any verdict, plus the two flags that stop a sort being wrong.
Decay Signatures
The five shapes a decline takes, what each one means, how to confirm it, and the route it goes to. Position decay is the rewrite. Presentation, demand, coverage and substitution are four other people's jobs.
Reading the Performance Data
Six things Google's performance report reference says about its own metrics that change what a register means, including why a link needs an impression for its position to be recorded and why an exported dash arrives as position 0.0.
Refresh Queue
The position-decay rows only, ordered by recoverable clicks rather than clicks lost, each carrying the queries the page stopped ranking for and the sections that are still working and must not be touched.
Control Set and Before and After Performance
Untouched pages in the same clusters, matched on pre-period clicks and chosen when the register is built rather than after the results are in, then the eight-week movement of both sets side by side.
Date and Markup Handling
What Google actually documents about byline dates, which is that they are its own estimate from several factors rather than a field anybody sets, and the one remedy it gives for a date that stays wrong.
Refresh Method and Update Log
How to turn a query diff into a specification, and a full worked cycle. The register, the routes, the edits, and the page that went backwards reported at the same size as the one that doubled.
How to use it
- 1
Open in River, or take it blank
Open the pack in River and send your performance history, or download the Word documents and CSV sheets and fill them in yourself.
- 2
Send pages and queries
Search performance by page and by query over the longest window you can pull, ideally past twelve months so the same weeks compare year on year, plus the articles or a content export.
- 3
Read the signature split first
The share of your lost clicks a rewrite can actually recover, and the largest single loss beside the work its signature calls for. Those two usually disagree, and that is the finding.
- 4
Refresh, then measure against the control
Body work on the queue, titles on a separate batch so the result attributes. Eight weeks later, both sets measured together, including the pages that fell.
Frequently asked questions
Is this template free?
Yes. The download is Word documents and CSV sheets, no account and no card. Edit with AI is the optional half: the agent runs your own performance history through the decomposition and fills the register. Every pack sits in the template library.
Why not just rank pages by traffic lost?
Because clicks are impressions multiplied by click-through rate, so that ranking is sorted by a number that already threw away the diagnosis. In the worked cycle the top two losers, 10,370 clicks between them, are a title job and a page nobody searches for any more.
Our top page's average position went up while traffic fell. What is that?
Decay, and the most reliable early signal there is. A link must get an impression for its position to be recorded, so a page that slips out of sight for its weaker queries stops averaging them in and the score improves across the survivors. The register flags every instance.
Does changing the published date help?
Not on its own. Google's byline date guidance describes the date shown in results as its own estimate drawn from several factors, so there is no field that sets it. Set the modified date because the page changed, and stop there.
A refreshed page still shows the old date in search results. What now?
Cut the other dates the template prints. Once the visible date and the structured value agree, Google's documented remedy for an incorrect date being selected is removing some or all of the other dates on the page: comment timestamps, related-post bylines, the sidebar of recent articles.
Why does the pack insist on a control set?
Because a batch that grows eleven per cent while the untouched pages in the same clusters grow fourteen has cost you three points, and no before-and-after on its own will ever say so. The control is chosen when the register is built, never after the results arrive.
How far back should the comparison window go?
Past twelve months, so the same weeks compare year on year instead of against last quarter. The performance report holds sixteen months, which is exactly one such comparison. Anything longer needs the API or a bulk export, and a refresh programme starts one on day one.
Find out which third of your decline is a rewrite
Take the Word documents and CSV sheets blank, or open this exact pack in River and send it the search performance history you already have.
Edit with AI