SERP tracking: where did your keyword ranking go?

It is 8:15 on a Thursday morning. At a mid-sized furniture retailer, the marketing lead starts the day by searching for the category page's target keyword. The page that sat in second place for three weeks now shows up ninth. A screenshot lands in the team channel with one line under it: "we lost the ranking." By midday three colleagues have run the same search from three laptops and seen three different answers - fourth, seventh, and not on the first page at all. In the afternoon meeting nobody can answer a single question with data: which day did the drop start, how many keywords moved, did a competitor climb or did the page slip?

This article is about the setup that makes that meeting possible. We cover, in order: why "position" is not a single number, the rules for freezing your measurement conditions, which keywords belong on a tracking list, a decision table for identifying which layer a drop came from, how to separate a competitor's rise from your own decline, why a drop without a date cannot be discussed, which movements deserve an alert, the distance between lost rankings and lost revenue, and who runs this work at what cadence.

When you say "the ranking dropped", whose screen are you talking about?

A search results page does not open the same way for everyone. The same query run from Manchester and from Bristol returns different results; mobile and desktop differ; a signed-in user's history shapes what they see; browser language, past searches and even the time of day shift the order. So the number one colleague sees on their own screen tells you nothing reliable about the site's actual visibility.

The second source of confusion is the shape of the results page itself. Tenth organic on a page of ten plain blue links sits just below the first screen. The same tenth position on a page topped by an ad block, a map pack, a featured snippet and a video carousel falls below the second scroll. Visibility can halve while the position never moves. The reverse happens too: when a block disappears from the page, fourth place suddenly starts earning far more clicks than it did the week before.

The practical consequence is that a position is not a measurement - it is a summary of one. The only number worth acting on is the curve formed by records taken repeatedly under identical conditions. A single day's screenshot is, at best, the start of a hypothesis. It is not grounds for a decision.

Why can't you argue about a drop before you fix your measurement?

The first rule of rank tracking is to write down the measurement conditions. At minimum, five things must be fixed: location (country and city), language, device type - mobile and desktop kept as separate records - the measurement frequency, and the source doing the measuring. Change any of the five and your records stop being comparable; what you are left with is noise in the shape of a chart.

The second rule is to keep the record raw, not interpreted. A note saying "we dropped, probably the update" is worthless six months later. A line reading "14 March, mobile, UK, 2 to 9" is not. Plenty of teams keep these records by hand in a spreadsheet and abandon it in the third month, because running searches, taking screenshots and filling cells every week is not sustainable work. The same reasoning behind the point at which a spreadsheet stops being the right home for tracking applies here without modification: the more often you have to record something, the less trustworthy a manual file becomes.

The third rule is a fixed interval. Checking five times a day does not produce better decisions than checking weekly; it just shows more wobble. The Keyword and SERP tracking capability inside Rocketly's Competitor Intelligence module does exactly this job: for every keyword you put under watch, the organic Google position is recorded on each collection run, so you end up with a series produced by one consistent method rather than a one-off glance. The refresh frequency setting scans busy competitors more often and quiet ones less often, and you can trigger a refresh by hand whenever you want.

Which keywords go on the list, and which ones stay off it?

A longer tracking list is not a better one. A hundred keywords guarantees that a handful moves every week, which produces constant alarms, which means nobody reads them by month two. A good list has three groups: transactional keywords that carry revenue, comparison keywords close to purchase intent, and brand keywords. Broad informational terms belong in the content plan, not the tracking list.

Building the list starts with separating intent, not with copying a search volume export. Track without knowing which keyword leads to a quote form and which one only leads to a read, and you will panic over the wrong term. Structuring keyword research around intent also gives the tracking list its skeleton: the intent groups you separate during research should survive as separate reports in tracking.

A common mistake at this stage is entering the same query with more than one page. When two of your own URLs take turns ranking for a term, the chart zigzags forever and you go looking outside for a cause that is sitting inside. Your own pages competing for the same query is the single most frequent cause of misdiagnosis in tracking data. When you add a keyword, write down the target URL alongside it; if nobody has decided which page should rank, nobody will notice when the ranking page changes.

Finally, build the habit of shrinking the list rather than growing it. A keyword that has not driven a single decision in three months comes off. A tracking list is not an archive - it is an attention budget.

Which layer did the drop come from?

A lost position rarely has one cause, but it almost always lives in one of four layers. Picking the right layer decides where most of your time goes: waiting out an "algorithm update" when the fix was a template change is expensive, and so is the reverse. The table below summarises the signal to look for in the first half hour and where to start.

Layer	Distinguishing signal	First step
Results page layout	Position unchanged, clicks down, a new block on top	Screenshot the results page and store it with a date
Your own page	Drop confined to one URL, aligned with a publish date	Trace the content, title and template change from that date
Competitor move	Your page is stable, a new domain has appeared above you	Put that domain under watch, record which page entered
Site-wide	Dozens of keywords start dropping on the same day	Check indexing and technical signals first
Seasonality	Position steady, overall search interest is down	Compare with the same period last year

What makes the table usable is that each row points at a different data source. The first looks at the results page, the second at your own publishing history, the third at a competitor's behaviour, the fourth at technical health and the fifth at demand. When those sources are not gathered on one screen, the team relitigates the same argument every time a chart moves.

Did the competitor climb, or did you slip?

Position is a relative measure. You may have gone from second to fourth because two new pages passed you, or because your own page weakened. Both produce the same number and demand completely different work. The only way to tell them apart is to measure yourself with the same method you use on competitors.

That is what the "your brand versus competitors" view in the Competitor Intelligence module exists for: your own brand is measured from the same sources as the competitors you follow, and the scores sit side by side in one chart. SEO strength monitoring records signals such as the number of indexed pages and the leading organic pages from one collection run to the next; if a rival published dozens of new pages this month, the dip on your keyword is not a coincidence.

A competitor's move is not always organic. Ad creative tracking shows the live creatives in the Meta, Google, LinkedIn and TikTok ad libraries along with the moments they changed, while news and search trends bring press coverage and search interest onto the same screen. If a rival published a new landing page and rewrote their ad headlines in the same week, the movement you see in the organic results has more than one author.

Reading a competitor's query strategy pays off on your paid side as well. Knowing which queries they show up for clarifies which queries you should not be paying for; the method for excluding queries that burn budget rests on the same logic as an organic tracking list - without relevance, visibility is just cost.

Why is a drop without a date impossible to discuss?

Most teams notice a drop without knowing when it began. Working from "it was fine a couple of weeks ago" makes every change made in those two weeks a suspect, and the conversation turns into defence rather than diagnosis. Once you know the day the decline started, the question narrows on its own: what shipped that day, which competitor entered that day, did the results page change that day?

A drop with no date has no cause - only opinions about it.

That is why the record should be built from snapshots rather than loose numbers. Historical snapshots keep every collection run, so you can pick two dates and replay the difference line by line. It lets you say "second place through 13 March, ninth from the 14 March collection onward", and the discussion concentrates on that day. Because the AI competitor summary compresses each run into a single readable paragraph, you can read what changed without walking through the numbers one by one.

Time-stamped record keeping is not an SEO-specific idea. On the sales side too, a measurement is worth something because you know when the event happened; recording when an email was opened and clicked is valuable for exactly that reason - it produces a timeline, not a number. In rank tracking the useful thing is not the position but the moment the position changed.

Which movements deserve an alert, and which are noise?

If every wobble becomes a notification, the notifications stop being read within three weeks. Setting up alerts means sorting movement by size and by the value of the keyword. The table below shows one way to draw that line; the numbers depend on the size of your list, but the logic holds.

Movement	Example threshold	Action
Transactional keyword falls off page one	Even a single keyword	Same-day alert, an owner is assigned
Shuffle inside the top three	Persists across two runs	Goes to the weekly review
One keyword moving in a broad list	Up to three positions	No alert, record it
Many keywords dropping on one day	A visible share of the list	Technical check, open a case
A new domain appearing in the top three	On first sighting	Add it as a tracked competitor

Alert thresholds and weights turn exactly this judgement into configuration: you set the score weights and the thresholds for your own organisation, and change detection only converts a movement into a notification once it crosses the threshold - and the notification goes to members with marketing permission, not to everyone. Small fluctuations are still recorded; they just do not interrupt anyone's day.

When you set thresholds, remember that what gets measured changes behaviour. A team that puts every keyword on a report will start optimising the easy ones within a month. The argument about what to measure and what to leave alone transfers directly: the measurement list is a description of the behaviour you are asking for.

Does a lost position always mean lost revenue?

No, and the distinction sits at the centre of most budget arguments. An informational page slipping from third to fifth may change nothing in revenue. A transactional page falling from second to fourth, on the other hand, loses more than the two places suggest, because clicks are distributed so steeply at the top. So a rank report is never read alone; sessions, form submissions and opportunity counts sit next to it.

The first requirement for that connection is knowing where the traffic came from. If campaign and source data are not carried through to the record, you cannot measure what an organic drop does to sales; carrying campaign parameters all the way into the record is the first pier of that bridge. The second is on the conversion side: conversion tracking set up correctly lets you read organic and paid with the same eye, and without it every dip gets papered over with ad spend.

The third pier is human. A request that arrives from organic search is won or lost according to how many hours pass before someone calls; the principles behind disciplined follow-up shift outcomes far faster than a ranking chart does. Gaining three positions takes weeks. Shortening the time to first contact changes the numbers inside the same week.

Who runs this, how often, and with which report?

As long as rank tracking stays one person's morning habit, it never becomes an organisational capability. A working setup needs three roles: someone who owns the list and the thresholds, someone who looks at alerts first, and a manager who runs the monthly review. Whoever receives the alert owns the case that follows; an unowned notification is an unread notification.

On cadence, one distinction saves a lot of time: collect daily, but do not meet daily. A weekly look sees movements; a monthly look sees the trend. A weekly email summary suits that rhythm - it carries the last seven days of movement to managers in one message and collects no new data, it reports what already exists. For anything going to a board or a client, report and PDF export writes the results to PDF or CSV and lets you put your own organisation's branding on the report header.

Finally, the ranking goal and the business goal belong in the same sentence. "Top three for ten keywords" is not a goal on its own; unless it is tied to an expected revenue outcome, the measurement is a vanity table. Setting a realistic target and tracking progress against it works the same way in marketing: a goal is a sentence with an owner and a review date. If you want your keyword list and your competitors on one screen, open a free account and start by putting your first five keywords under watch.

FAQ

How often should rankings be measured?

A daily measurement is enough for most businesses, and consistency matters far more than frequency - the same conditions every time. Fast-moving sectors justify more frequent runs, quiet ones fewer. Keep the meeting weekly, though; deciding on daily wobble is how teams end up rewriting pages that were never the problem.

Why does the position on my screen differ from the tracking tool?

Because your own search is personalised: your location, search history, signed-in state and device all shape it. A tracking tool measures against a fixed location, language and device definition. What is comparable is the record taken under fixed conditions; your own screen is a spot check, not a data point.

How many keywords should I track?

Tie the list to decisions: if movement on a keyword leads to no action, it should not be tracked. For most teams a short core list of revenue-carrying terms produces more work than an archive of several hundred. Reviewing the list at the start of each quarter and removing what nobody used is a good habit.

What should I do when a new domain enters my competitive set?

First record which page it entered with and on which keywords, then add it to tracking. A newcomer appearing on one keyword can be an accident; appearing across several in a row is a strategy. Telling them apart means logging the first sighting with its date and comparing it against the next run.