The settings decide the number before your SEO does
A rank tracker is not a window onto reality. It is a robot that runs a search from a configured location on a configured device and writes down what it saw. Change the configuration and the number changes, with nothing about your site having moved at all.
Google itself is explicit that results vary by time, location, language, device and personalisation. Two people searching the same phrase minutes apart in the same city can see different pages. Your tracker picks one point in that space and reports it as though it were the position — which is fine, as long as you chose the point deliberately.
These are the settings that matter, roughly in order of how much damage the wrong choice does.
| Setting | What goes wrong at the default | What to set |
|---|---|---|
| Location granularity | Country-level tracking for a business that sells in one city. Commercial queries in India vary heavily between Bengaluru and Patna, so the national number describes nobody. | City level, for every city you actually sell in. One national series only if you genuinely sell nationally. |
| Device | Desktop-only tracking on a market that is overwhelmingly mobile. The two result pages differ in layout, in how much is above the fold, and often in which pages rank. | Mobile as the primary series. Desktop as a second series, not an average of the two. |
| Language and country domain | Tracking google.com in English for a business whose buyers search on google.co.in, sometimes in Hinglish or a transliterated term. | The country domain and interface language your buyers use. Track regional-language variants as their own keywords. |
| Personalisation and login state | A logged-in or history-carrying session flatters you, because Google has learned you keep clicking your own site. | Logged-out, un-personalised. Any credible tracker does this; verify it rather than assuming. |
| SERP features counted | Some tools count the local pack, an AI Overview or a video carousel as a position; some do not. Switch tools and every number shifts. | Organic-only as the headline series, with features tracked as presence or absence. |
| Check frequency | Daily checks on a low-volume keyword set produce a chart made almost entirely of noise. | Weekly for most sites. Daily only where the query is volatile and the money is large. |
The uncomfortable bit about how the data is collected
Worth knowing before you buy anything. Rank trackers work by issuing automated queries against Google at scale, and Google's own spam policies class scraping results for rank-checking purposes as machine-generated traffic conducted without permission.
In practice this is a rule Google does not enforce against the buyer of a tracking tool, and an entire industry runs on it. But it explains a few things that otherwise look like bugs: why tools occasionally miss days, why results can differ between two tools on the same morning, and why nobody can promise you a complete or authoritative dataset.
It also explains why Search Console is the more defensible number. It comes from Google, it covers every query your real audience searched rather than a list you guessed at, and it is free. Its weakness is that it reports a period average across all locations rather than a moment in one — which is exactly the weakness a tracker fixes. Use both, and understand which is which; average position explains why the Search Console figure behaves the way it does.
Forty keywords chosen by revenue, not four hundred chosen by volume
The default instinct is to track everything, because the tool prices by keyword and 500 feels thorough. It is the opposite of thorough. A 500-row report cannot be read, so it does not get read, so it becomes a monthly attachment nobody opens.
Our working number is around 40 for most businesses, and it is chosen on a different axis entirely.
- Start from revenue, not from a keyword tool. List the five to eight things you sell that actually pay salaries. Every tracked keyword has to belong to one of them or it does not go on the list.
- Take the two or three queries a buyer types when they are ready. These are usually unglamorous, medium-volume and specific. They convert, which is the only property that matters here.
- Add the queries you already rank 5–20 for, from Search Console. These are the movable ones. A keyword you rank 90th for tells you nothing weekly for the next six months.
- Add three or four brand queries, tracked separately and never mixed into the average. Brand positions move for reasons unrelated to your SEO work and they will flatter every report you put them in.
- Add up to five you are deliberately chasing and cannot currently win. Label them as ambitions in the report so nobody mistakes a flat line for failure.
- Stop. If the list runs past 60, you are tracking a keyword research exercise rather than a business.
Map results are a different instrument
If your buyers search "dentist near me", the thing you care about is the map pack, and a standard rank tracker handles it badly or not at all.
The reason is that local results are ranked on relevance, distance and prominence, and distance is measured from wherever the searcher is standing. There is no single position. A clinic can sit first for someone two streets away and be absent for someone four kilometres north, at the same instant. A tracker configured to "Pune" resolves that to one coordinate — usually the city centroid — and reports a number that only describes people standing there.
Three honest options, in ascending order of cost:
- Track it as presence, not position. In the pack or not, for each target query, from your own location. Crude, free, and enough for most single-location businesses.
- Grid tracking. Tools that check the pack from a lattice of coordinates around your address and return a heat map. This is the correct instrument, and it is the only one that shows how far your visibility actually reaches.
- Track it not at all, and report Google Business Profile calls, direction requests and website clicks instead. For a business with one location and a phone that rings, this is often the most honest choice available.
- Never mix pack positions into your organic average. They are separate result sets with separate rules, and combining them produces a number with no meaning at all.
Never report a position without its clicks
Position is an input. Clicks are the outcome. Reported alone, position invites exactly the wrong conclusion about half the time, because the result page keeps changing shape underneath it — an AI Overview appears, a shopping row lands, four ads push you below the fold while your reported position stays at 4.
Put the two side by side and every combination tells you something specific.
- Pull the clicks from Search Console, filtered to the same query and the same country as the tracked row. Anything else is comparing two different populations.
- Show the 28-day comparison, not the single week. Weekly noise on a medium-volume keyword routinely swings three positions with nothing behind it.
- If the two disagree for two consecutive months, believe the clicks. See keyword rankings vs organic traffic for why they diverge more often than they used to.
| Position | Clicks | What it usually means |
|---|---|---|
| Improving | Rising | The work is landing. Nothing to investigate. |
| Improving | Flat or falling | The result page changed shape, or the query lost demand. Open it in an incognito window before you touch the page. |
| Flat | Rising | Usually a title or snippet improvement, or you have started ranking for more long-tail variants of the same theme. |
| Falling | Flat | Often a competitor entering above you on a query that was never the one sending your clicks. Check which query moved before reacting. |
| Falling | Falling | The only combination that is unambiguously bad, and the only one worth an emergency. |
The reports we refuse to send
Rank tracking is easy to weaponise, which is why so much agency reporting is built on it. These are the four we will not produce, and the reason in each case is the same: they are technically accurate and designed to mislead.
None of it is fraud. It is selection, which is harder to argue with and does more damage.
- The 500-keyword table with the green arrows. Track enough keywords and some always improve. A report that shows only the movers is a report engineered to be positive in every possible month.
- Sitewide average position as a headline. It gets worse when a site starts ranking for more things, which is to say it gets worse when the work is going well. It belongs in no summary.
- Positions with no baseline date. "Ranking 4th for kitchen interiors Bangalore" means nothing without what it was in January and what it was measured on.
- Brand keywords folded into the average. You already rank first for your own name. Including it lifts the average and reports nothing about the work you are paying for.
- Rankings without a lead count anywhere in the document. Positions are a leading indicator. The number the invoice should be argued against is enquiries, which is why our own guarantee is written against a client's frozen trailing-90-day organic lead baseline rather than any position at all.