Improve existing pages or publish new articles?

Most content plans default to new articles. A calendar of fresh posts is easy to plan and easy to report, and it feels like progress. Meanwhile the pages that already show up in Google, and already bring in customers, sit untouched for years.

The opposite mistake happens too. A team rewrites the same handful of pages every quarter while the searches its customers actually make have no page at all.

Both come from deciding by habit instead of evidence. The evidence is already in Search Console and your analytics: how old each page is, how much demand it already sees, whether it converts, and whether another of your pages is competing for the same search. This guide turns those four signals into a decision table, then runs three pages through it.

Four signals that decide it

1. Page maturity

New pages move around. Google has to crawl a page, index it and work out where it belongs, and positions in the first weeks often swing. Judge a page too early and you will rewrite something that was still settling, then never know whether the rewrite helped.

As a working rule, leave a page alone for about three months after publishing before deciding it has underperformed. That is a rule of thumb, not a Google figure: a large site that Google crawls every day settles faster, a small new site slower. What you are waiting for is a page whose average position has been roughly stable for several weeks. That is a page you can judge.

2. Existing demand

Search Console's performance report shows, for each page, the searches it appeared for, its impressions (how many times it was shown in results) and its average position. Impressions are measured, not estimated, so they are the best evidence you have that people want what the page offers.

The most promising pages have real impressions at positions 4 to 20. Google already considers them relevant, but they sit low enough that a move up changes the clicks a lot. Using typical click-through rates (the share of searchers who click a result, which varies from one results page to another), a result at #9 gets about 2.1% of clicks and a result at #5 about 5.2%. That move roughly multiplies the page's clicks by two and a half, and nothing new has to be written.

A new article starts with no impressions and has to earn its place from nothing.

3. Conversion evidence

Traffic is not the goal. In Google Analytics, or whatever analytics you use, check whether visitors who arrive on the page from search go on to buy, sign up or enquire. A page that already converts has proved that it matches what those searchers want, so more of the same visitors will probably convert at a similar rate.

A page with plenty of clicks and no conversions is usually answering a learning search (someone who wants to know something, not buy something). Improving its ranking adds traffic, not revenue. That page's job is to pass readers to the pages that sell.

4. Cannibalisation

Cannibalisation is when two or more of your own pages compete for the same search. To spot it, open one query in Search Console's performance report and switch to the Pages tab. If two of your URLs share its impressions, or the page Google shows keeps changing from week to week, you have it. Google usually shows one of them at a time, and two half-relevant pages tend to do worse than one clearly relevant page.

This is the signal that most often rules out a new article. If the search you plan to write for already has one of your pages in the top 20, a new page on the same topic creates a competitor for it.

The decision table

Check the signals in this order: maturity first, then cannibalisation, then demand and conversion. A young page or a duplicated topic changes what the other numbers mean.

A worked example: three pages

Take a company selling booking software to hair salons. The figures are illustrative. Search figures are from Search Console for the last 28 days; trials are organic visits that started a free trial, from Google Analytics.

Page A: improve it

It is mature, its main search is a buying search, and it sits at position 9.2 with 20,000 impressions. The clicks match the typical figure for #9: 20,000 × 2.1% = 420. It converts: 17 trials from 420 clicks is 4.0%.

What would a realistic move be worth? Not #1, which would mean overtaking the strongest pages in the market, but #5:

• Clicks at #5: 20,000 × 5.2% = 1,040
• Extra clicks: 1,040 − 420 = 620 a month
• Extra trials at the same rate: 620 × 4% = 24.8, about 25 a month
• If one trial in four becomes a customer worth $600 in the first year, a trial is worth 0.25 × $600 = $150
• Modelled value: 24.8 × $150 = $3,720 of first-year revenue for each month at the new position

There is one complication: an old post, /blog/choosing-salon-software, also appears for "salon booking software" at position 17. That is cannibalisation. Before rewriting Page A, move the post's useful sections (it has a good checklist) into Page A and redirect the post there.

Page B: wait

Six weeks is too early. Its position of 23.4 tells you very little yet, and it is a learning search that will never produce many trials directly. The one change worth making now is small and doesn't disturb the page: a link from the post to Page A, placed in a sentence about reducing no-shows with deposits and reminders. Re-check the post at three months.

Page C: publish it

The team's first idea was to add a section on mobile hairdressers to Page A. Search Console argues against it. Page A gets no impressions for these searches, and the only page of theirs that appears is the homepage, at position 47, for 60 impressions. Mobile hairdressers have different needs (travel time between clients, no front desk), so the searches have a different intent from Page A's main search. A separate page won't compete with Page A because Page A doesn't rank for them.

There is no Search Console demand to measure here, so treat any volume figure from a keyword tool as an estimate. The sales team's evidence that mobile hairdressers sign up and stay is the conversion signal.

The order of work

Page A first: the demand is proven, the conversion rate is known and the upside can be modelled. Page C second. Page B waits. A content calendar built on habit would have done these the other way round.

Ship/Scale does the Page A part of this for every page with impressions in Search Console: it ranks pages by the revenue a move up would add, using each search's intent and monthly volume, ignores brand searches, and takes the value of a click from your connected analytics or from figures you set.

Where this goes wrong

• Rewriting a page that is already winning. A page in the top three has little to gain and a lot to lose. If you must change it, change one thing and measure it against the weeks before. The guide to measuring SEO changes at 30, 60 and 90 days covers how.
• Impressions on page two understate demand. Beyond position 10, an impression only counts when someone loads the results page your page is on, and most searchers never go past the first. A page at position 14 may have far more demand behind it than its impressions show. Where you have a monthly search volume, use it alongside impressions.
• Average position hides a spread. A page's average blends every search it appeared for. "Position 9.2" can mean #3 for one search and #40 for another. Look at the page's main searches one by one, not only the page total.
• Small conversion numbers are noisy. Seventeen trials is not much. A month with 12 and a month with 22 can both be chance. Use three months of data where you can before trusting a conversion rate.
• Merging loses something. A redirected page loses the searches that only it answered. Before merging, list the old page's queries in Search Console and carry over the sections that earn them.
• New markets have no history. If you are entering a category you have never written about, Search Console has nothing to show you. Publishing is then the only option, and the estimates stay estimates until the page has three months of data.

Once a change is live, Ship/Scale measures it against the 28 days before it at 7, 14, 30, 60 and 90 days (position, clicks and impressions for the searches it targeted), with forecasts labelled as modelled and results as measured. That is how you find out whether "improve" was the right call.

The short version

Check the signals in order: leave pages under about three months old alone, merge pages that compete for the same search, then look at demand and conversion. Mature pages with buying searches at positions 4 to 20 and proven conversions should be improved first, because the demand is measured and a realistic move up is worth estimating. Publish new pages for searches none of your pages answer. Judge both by measured results, not by how much was published.