Rank tracking is a measurement. Measurements do not edit pages. If you are asking why track keywords at all, the honest answer is that tracking only earns its cost when every number on the dashboard has a named action attached to it, and a person who owns that action. Otherwise you are paying a monthly fee to feel informed.

Most lean teams we see have the tracker running. The weekly email arrives, someone scans it on a Tuesday morning, a few positions are green, a few are red, and nothing on the website changes. Three months later the same forty keywords are being tracked and the same pages are live, word for word.

That is the failure mode worth writing about. Not "pick the right keywords". Not "align content with intent". The gap between the number moving and anybody doing anything.

The report nobody acts on

Tracking is a measurement layer. A publishing workflow is a different thing entirely, and in most small teams it does not exist as a written process: there is no step that says "position 14 with impressions means rewrite this page by the end of the month".

So what does the tracker actually buy you? A sampled view of where your URLs sit for terms you chose, over time. That is useful. It tells you whether an edit worked, it catches a page falling off a cliff, and it shows you which terms you are close on. What it does not buy is the decision. The decision is the expensive part, because it costs attention rather than money.

Here is the test to apply to every keyword in your list, right now. Can you name, in one sentence, the action that a movement in this keyword would trigger? If the answer is "I'd have a look", that keyword is a cost with no return. Delete it or assign it.

Forty tracked terms with no owner is worse than eight with one. We would rather a plumbing firm tracked eight terms it genuinely competes on and edited two pages a month than tracked forty and edited nothing.

What keyword tracking honestly tells you, and what it cannot

A position is one sampled view of one result page. It does not tell you whether the searcher was ready to buy, whether the term carries any margin, or whether intent has shifted under your feet. A term can hold position 3 for a year while the people searching it change completely.

Average position in Search Console is blended by default. Google's own documentation on the Performance report explains that position is taken from the topmost result for your site and averaged across the queries, devices and locations in the filtered set. Change the mix of queries that generate impressions, and your average moves without a single page getting worse. That is not a ranking drop. That is arithmetic.

SERP features change the value of a position overnight. Position 3 above the fold in 2019 and position 3 under an AI overview, a People Also Ask block and a local pack are not the same asset. Your rank tracker will usually report both as "3". Google has been expanding AI Overviews across UK results since 2024, and the practical effect for publishers is that organic position and screen position have come apart.

Which is why impressions and click-through rate are the earlier signal. Both tend to move before rank does, and both respond to things rank cannot see: a new feature above you, a competitor rewriting their title tag, a seasonal shift in how people phrase the query. If you only look at position, you find out last.

Five rules for turning movements into decisions

These are the rules we use to convert a tracker row into a brief. Each one ends in a named action, including the one that ends in nothing.

1. Positions 11 to 20 with real impressions: rewrite, do not republish

Striking distance is the cheapest work on the list. You already have a page Google considers relevant enough to show on page two, and you already have impression data telling you which query variants it is getting shown for. Rewrite that page against the live top 10: match the format people are actually being shown (guide, comparison, list, local service page), answer the query in the first paragraph, and expand the sections competitors cover that you skipped. Publishing a second page on the same topic instead is the single most common mistake here, and it causes problem three below.

2. Stable rank, falling click-through rate: that is a title job

A page holds position 4 for twelve weeks and clicks halve. Nothing on the page changed, so nothing about the page failed. The SERP changed around it: a feature appeared, or somebody above you rewrote their title into something more specific. The fix is a sharper title, a meta description that matches what the searcher wants, and a first paragraph that answers the question outright rather than warming up to it. Do not commission a new article for this. It is an hour of editing.

3. Two of your URLs swapping places for one query: cannibalisation

This is the most commonly misread row in rank data. It looks like volatility, because the number jumps around week to week. It is usually two of your own posts competing for the same intent, neither settling above position 8, with Google alternating between them because neither is clearly the better answer.

Merge them. Take the stronger URL, fold in whatever the weaker one did better, give it one clear target, and redirect the loser. Prevention is cheaper: check your existing pages and previously published articles before anything new is briefed, which is exactly why a research-first process beats a content calendar full of titles somebody liked the sound of. We cover the mechanics of that check in our full content workflow.

4. Zero impressions after eight to twelve weeks: wrong target, not bad luck

Impressions are a low bar. If a page has had none for three months, Google is not withholding traffic out of caution, it has decided your page is not a candidate answer. Usually the page type is wrong for the query: a blog post aimed at a term that returns product pages, or a service page aimed at a term that returns definitions.

Take the plumbing example. Forty tracked terms, thirty flat at zero impressions after three months, five sitting at 11 to 20 with real impressions. Those five are the whole month's work. The thirty should come off the tracker, because a flat line produces no information and still occupies your Tuesday morning.

5. A wobble of a few positions: do nothing

Position 6 to 8 and back to 6 is sampling noise. Rewriting a page on that basis is how teams end up rewriting twice and learning nothing from either edit. Set a threshold before you start, hold to it, and resist the urge to act inside it.

Before you track: the research that makes the numbers mean something

Tracking a keyword you never had a realistic shot at gives you a flat line and a false sense of activity. The research happens first, or the tracking is theatre.

Two things matter more than list length. First, UK-specific volume and difficulty: global figures flatter terms nobody here searches, and an SME chasing a term dominated by national retailers is funding a dashboard row, not a pipeline. Second, freshness. Difficulty pulled six months ago does not tell you who is on page one today, which is why our weekly live keyword scan refreshes UK volume and difficulty every week rather than once at onboarding.

Then run a gap check against a handful of competitors, five to ten is plenty, and look for terms they rank for and you do not. That is where the genuinely winnable briefs come from. We have written separately on how competitor keyword tracking surfaces content gaps, and the short version is that the gap list is a better starting point than a brainstorm.

Last step before anything is written: check what you have already published. Duplicate titles split the data and set up the swap described above.

False alarms: when the number moved but your content did not fail

Mobile and desktop results differ, sometimes substantially, and a blended average hides which one slipped. Split the device filter before you conclude anything, or you will rewrite a page that is fine on desktop and broken on a phone for reasons that have nothing to do with the copy. We go into the detail in our piece on whether your mobile rankings are really slipping.

Crawl depth quietly delays the whole feedback loop. A page sitting four clicks from the homepage, reachable only through pagination, gets re-crawled less often than one linked from a main section page. You edit it on Monday, you check the tracker a fortnight later, nothing has moved, and you conclude the rewrite failed. It may simply not have been seen yet. Check the last crawl date in Search Console's URL Inspection tool before you judge any edit.

Then the ordinary stuff that looks like decline: seasonality, a shift in the brand versus non-brand mix, and one high-volume query dragging a whole folder's average around. And volatility during an unconfirmed update, where the sensible move is to wait for a stable baseline rather than rewriting into a moving target. Google confirms core updates publicly; if the industry is twitchy and Google has said nothing, give it a fortnight.

A monthly review that ends in a to-do list

The cadence that works for small teams: refresh data weekly, make decisions monthly, audit properly each quarter. Weekly looking with monthly deciding keeps you from reacting to noise while still catching real falls.

At the monthly review, every tracked keyword goes into one of four buckets, and nothing is allowed to sit between them:

• Rewrite: page two with impressions, or stable rank with falling clicks.

• Consolidate: two URLs trading places for one query.

• Build new: a genuine gap where no page of yours is a candidate.

• Retire or ignore: zero impressions after three months, or movement inside your noise threshold.

Done honestly, a quarterly audit usually produces a handful of pages to work on, not dozens. That is the point. A monthly decision rhythm beats a weekly panic, and it stops the pattern where a bad report triggers a dump of ten thin new pages that split your own data further.

Log the date of every change. Title rewritten 12 March, two posts merged 4 April, internal links added 19 April. Without the log, next month's review cannot attribute any movement to any action, and you are back to guessing. The log is also what tells you honestly when your own edit did nothing, which is information most teams never collect. Our approach to research and briefs starts from the same place: the brief is planned against the live top 10 for the term, so the next step after a movement is already named.

Keep publishing at a steady rate rather than in bursts. Gradual beats lumpy, for crawl reasons and for your own sanity.

So, why track keywords? To trigger edits. If the last three months of tracking produced no rewrite, no merge and no new page, the tracker is not the problem and neither is Google. The missing piece is the decision step, and it costs nothing to add except the discipline of writing the action down next to the number.

Frequently Asked Questions

Why track keywords if rankings move every week anyway?

Because the weekly wobble is not the signal, the trend and the surrounding metrics are. Tracking tells you which pages are within reach of page one, which have lost clicks at a steady position, and which never got shown at all. Those three findings each map to a different edit. If you cannot name the edit, stop tracking that term.

How often should a small team review keyword tracking data?

Refresh the data weekly, but make decisions monthly. Weekly decisions mean reacting to sampling noise and rewriting pages twice. A quarterly audit on top of that, where every term is sorted into rewrite, consolidate, build new or ignore, is usually enough to keep the to-do list short and real.

Is Google Search Console enough, or do I need a rank tracker?

Search Console is free, uses your own data, and gives you impressions and click-through rate, which are the earlier signals. Its average position is blended across devices, locations and query variants, so filter before you draw conclusions. A rank tracker adds daily positions for specific terms and locations, and visibility of SERP features; worth it once you are actively editing pages each month, hard to justify if you are not.

How many keywords should an SME track?

Fewer than you think. A tracked keyword needs an owner and a named action, so the practical limit is how many decisions you can actually make in a month. Ten to thirty terms that map to real pages and real commercial intent beats a list of hundreds you skim and forget.

What rank movement is big enough to act on?

Set the threshold in advance rather than judging each case. Movement inside a few positions, especially in the teens and twenties, is normally noise. A sustained fall over three or four weeks, a page dropping off page one, or a stable position with clicks halving are all worth investigating, and only the last of those is likely to be a title problem rather than a content one.