serps.io
Guides
Beginner·6 min read

How to set up rank tracking the right way

Tie every keyword to a real page, match location and device to your market, and learn which movement is signal and which is noise.

Decide what the tracker is for before you add a keyword

A rank tracker answers one of two questions, and the setup differs. Either you are proving that a specific page is gaining or losing ground, or you are watching a market so you notice when a competitor moves. The first wants a small list checked often, with each keyword tied to a URL you can edit. The second wants wider coverage, checked less often, grouped by competitor rather than by page. Pick one before you import anything. The failure mode is a tracker holding every term anyone ever mentioned, so nobody opens it: four thousand rows all drifting slightly, with no way to tell which drift matters. If two people need different answers, build two projects rather than one list that serves neither well. Write the question at the top of the project description so whoever inherits it later knows what it was built to answer. Ten rows someone reads every Monday beat a thousand that run on autopilot. If you inherit an existing tracker, assume the question was never written down, pick one yourself, and archive the rows that do not serve it before adding anything new.

Pick keywords that map to a page you own

For every keyword you add, name the URL you expect to rank. If you cannot name one, you are not tracking a ranking, you are tracking a wish, and the row will sit near the bottom of page eight forever while dragging down your averages. Keep a column for the target URL and compare it against the URL that actually ranks. Three things surface quickly: keywords where a blog post ranks and the product page should, keywords where two of your own pages alternate week to week, and keywords where a page you had forgotten about is doing the work. All three are fixable, and none are visible from position alone. Drop terms that only your homepage could ever satisfy, since homepage rankings move for reasons you cannot trace back to any single change. Add those terms back once a real page exists for them, not before.

Set location, language, and device to match your buyers

Position is a function of where the query was made. A term can sit at four nationally and at eleven in the one city that drives half your revenue, and a national check will never show you that. Configure each project with the country, the language, and where it matters the city or postal code you actually sell into. For a service business with physical locations, run one project per location rather than averaging them, because an average hides the branch that is losing. Set device separately whenever mobile and desktop results differ, which they usually do for anything with a local pack or a shopping carousel. The common mistake is leaving country and language at defaults that match your office rather than your market, then being surprised when reported positions do not match the screenshot a customer emails you. When you expand into a new market, add the project before the launch rather than after, so you have a before picture to compare against. Checking a new country from your existing project and seeing nothing is not evidence of anything except a misconfigured location.

Choose a check frequency you can act on

Daily checks feel responsive and mostly produce noise. Normal day-to-day variance for a mid-page term is a position or two, and reacting to that is how teams end up rewriting the same page every Tuesday. Weekly is the right default for most sets: it smooths the jitter and still catches a real drop within days. Check daily only for the handful of terms where a drop costs money immediately, such as your highest-converting commercial pages, or during the week after a migration or a known algorithm update. Keep those two groups separate so the daily list stays short enough to read. If you are tempted to check hourly, be honest about what you would do differently with the extra data. Usually nothing, because you cannot diagnose a cause and ship a fix inside an hour anyway.

Record the ranking URL and the result layout

Position alone hides the two changes that explain most traffic swings. The first is a URL swap: you were at six with a detailed guide, now you are at six with a thin category page, and clicks collapsed because the intent no longer matches. The second is a layout change: you held position three, but an AI answer, a video carousel, and four ads appeared above you, so position three now sits below the fold. Capture both. If your tracker stores the ranking URL for each check, diff it weekly and treat every change as an event worth ten minutes. If it records which features appeared on the page, watch for the first appearance of a local pack or an answer block on your money terms, because that is usually the exact moment a stable keyword stops sending the traffic it used to send. Screenshot the results page for your five most valuable terms once a month and keep the images. When someone asks why clicks fell while rankings held, a pair of screenshots three months apart answers it in seconds and no table will.

Set a baseline, then annotate everything you ship

Rankings move for reasons you caused and reasons you did not, and a month later nobody remembers which was which. Fix the date you started tracking, let two or three weeks pass without touching anything, and treat that as your baseline. From then on, log every change against a date: page rewrites, title changes, a redirect, a new internal link block, a migration, a CDN change, a known algorithm update. The log is the whole value. Without it, a drop in week six is a mystery. With it, you can see the drop began the day a template change rewrote the H1 on two hundred pages. Keep the log in one place, keep it boring, and include changes nobody considers SEO work, since a developer swapping how a section renders is exactly the kind of thing that moves rankings and never gets mentioned in a standup.

Read the trend, and define what a real drop looks like

Decide in advance what counts as a signal, then ignore everything else. A workable rule: a move of fewer than three positions in a week is noise, a move of five or more held across two consecutive checks is real, and a keyword leaving the top ten entirely is urgent regardless of size. Sort by change rather than by position, because the terms already sitting at the top of page one are the least likely to need your attention. When something does drop, check in this order: did the ranking URL change, is the page still indexable, did a competitor publish something new, did the result layout change, and did you ship anything that week. Most drops resolve to one of those five. If none of them explain it and similar keywords moved the same way, you are probably looking at an algorithm update, and the right response is to wait a week before rewriting anything.

Prune and re-scope every quarter

A keyword list rots. Products get renamed, pages get merged, campaign terms stop mattering, and demand drifts. Once a quarter, remove anything that has sat outside the top fifty for two quarters with no page shipped against it, anything whose target URL now returns a 404 or a redirect, and anything nobody has looked at since it was added. Then add the terms you have started appearing for without trying, which you can pull straight from your search analytics. Those are the cheapest wins available, because the page already exists and already half ranks. Keep the list small enough that one person can read all of it in a sitting and still care about each row. A pruned tracker is a tool. A tracker that only ever grows becomes a report nobody opens, which costs the same as not having one at all.