How to set up rank tracking for a topic cluster
Tag the cluster's queries as one group, read its median position and top-10 count every week, and stop rewriting pages because one keyword moved.
Rank tracking for a topic cluster works at the group level. Tag every query the cluster targets as one segment, then read two numbers a week: the median position of the group, and how many of its keywords sit in the top 10. One keyword moving six spots tells you almost nothing. The group line tells you whether the cluster is gaining ground.
Why one keyword can't tell you how a cluster is doing
Search Console's average position is an average across queries. Google's documentation says the number shown is “the topmost position occupied by a link to your property or page in search results, averaged across all queries in which your property appeared”, and it warns that “a position number can mean different things in different situations, and so you should not make simple assumptions”.
Apply that to a cluster. Your pillar page slips from 7 to 11 on its head term. In the same week, four supporting pages climb into the top 20 on queries the pillar never ranked for. Read the head term alone and you had a bad week. Read the group and you had a good one.
Tracker cadence adds a second layer of noise. Ahrefs Rank Tracker updates tracked keywords weekly by default on every subscription plan. Semrush Position Tracking updates daily, generally within 24 to 48 hours, and keeps daily points for 60 days before it switches to weekly ones. Open a weekly tracker on Tuesday and again on Friday and you are reading the same crawl twice, then explaining a change that never happened.
How to set up rank tracking for a topic cluster
Five steps. Budget half an hour for a cluster of ten pages.
- Write the query list in one place. The pillar's head term, one primary query per supporting page, and the variants Search Console already shows for those URLs. Twenty to sixty queries is a workable cluster.
- Give them one tag. Ahrefs, Semrush and most trackers let you tag keywords and filter every report by that tag. One tag per cluster, named after the pillar. Hold off on sub-tags until the cluster passes thirty pages.
- Point the tag at a URL folder. When the cluster's pages sit under one path, you can pull the same set out of Search Console with a page filter and cross-check what the tracker claims. This is the practical argument for grouping cluster pages into their own URL folder.
- Record the baseline before the next page ships. Date, median position, top-10 count, folder impressions. Four cells in a spreadsheet row. Skip this and every later reading is a guess.
- Match the cadence to your publishing pace. Publishing weekly, read weekly. Publishing monthly, read monthly and keep the weekly data for context. Daily checks on a cluster fill meetings and settle nothing.
The three lines that carry the story
- Median position of the tagged group. Median, so one page stuck at 90 doesn't drag the whole picture down with it.
- Keywords in the top 10. Count them. The number moves slowly, which is what makes it readable.
- Impressions for the folder in Search Console. This catches demand you never thought to track. A cluster gaining impressions on untracked queries is working, whatever the tracker shows.
Three things to leave alone. The daily line for any single keyword. Any one-week move under five positions on one query. And the tracker's estimated traffic column, which multiplies a position guess by a volume guess and gives you a number with two errors baked in.
When a move is worth acting on
One keyword drops ten spots: write it down, do nothing. The whole tagged group drops five positions in a week: something changed, and the cause is usually outside your site. Two pages in the cluster trading places on the same query week after week: that one is yours, and it means the two pages are competing for the same intent.
Before you rewrite anything, check what you published in the two weeks before the drop. The most common self-inflicted cause is a new page that competes with the pillar instead of feeding it, which is a planning problem more than a writing one. Our guide to building a topic cluster that ranks covers how to assign one query to one page before you brief it.
Where the setup usually breaks
Search Console truncates report exports to 1,000 rows of representative examples. A mature cluster with a long tail passes that in the interface without warning you, so filter to the folder before you export, or pull the query through the API when you need the full set.
The quieter failure is drift. Pages get added to the cluster and never added to the tag, so the group line flattens while the cluster grows. Put the tag in your publishing checklist next to the internal link step, and re-run the query list every quarter against what's actually in the folder.
None of this needs a new tool. A tracker you already pay for, one tag, one folder filter in Search Console, and four numbers a week will tell you more about a cluster than a dashboard nobody reads.
If you want a cluster set up once and read properly every week after that, that's the kind of work we do. Tell us which cluster you're building and we'll show you what we would track.
