How Often to Check Rankings? Twenty Queries, Twice

How it was measured

Twenty Persian queries, all from markets we actually work in: hosting, SEO, web design, payment gateways, WordPress security. Google organic results to depth 20 for each, location code 20498, language fa. Seven queries had roughly a day between snapshots, six had about three days, seven had about four.

The comparison is URL against URL, not domain against domain. That distinction turns out to matter, and there is a section on it below.

199 top-ten placements were tracked in total and 175 of those URLs were still in the top ten at the second reading.

Here is the part that decides the argument. Of those 175, 89 had not moved a single place. Forty-one had gone up and forty-five had gone down. That split is what random looks like. If some real force were reshuffling these results, one direction would be heavier. Neither is.

The five cases only a URL comparison catches

Five times, the domain held its place in the top ten while Google quietly swapped which page of that domain it was ranking. On the query «درگاه پرداخت», zibal.ir sat at position 5; three days later zibal.ir was still in the top ten, at position 6, with a different URL.

A tracker that counts domains calls that no change. A tracker that counts URLs calls it one exit and one entry. Neither is lying, and they are answering different questions. When a rank report suddenly looks strange, check which of the two it counts before you check anything else.

One query refused to behave

«WordPress security» in Persian turned over four of its ten results in under three days. Faradars fell from position 2 clean out of the top ten. Maktabkhooneh and Daneshjooyar went with it. Rtl-theme, which had been fifth, came back third on a different URL. Zhaket, YouTube and a hosting company took the empty seats.

Over the same window, «cheap web design», «free website builder» and «SEO company» in Persian barely twitched. Volatility is not a property of Google. It is a property of the query. A result set built from tutorials and recent blog posts churns; a result set built from settled commercial pages does not.

What changed fastest was not the rankings

In those same four days the query «content production» gained a featured snippet it did not have, awarded to college.tapsell.ir. «Google advertising» lost the featured snippet Yektanet had been holding. «SSL certificate» lost its image pack.

So in a window where eight results in ten stayed put, the shape of the page changed three times. If organic traffic dropped while your position did not, this is the first place to look. A snippet appearing above you costs clicks without costing you a single place.

So how often should you look?

Our answer, offered as an opinion rather than a finding: weekly for reporting, daily only when you have a specific reason. A specific reason is a large publication, a domain or URL migration, hours of downtime, or an update Google has announced. In those cases you are watching for one identified event, and daily makes sense.

Outside them, a daily number wanders between minus two and plus two and you are buying yourself anxiety. One rule follows directly from the data: do not report a one or two place move at all. Across the one-day interval the median move was zero and the largest was five places, ayratech.org going 9 to 4 on a Persian query for bespoke web design. Anything inside the noise band does not belong in a client report.

What this does not prove

• Twenty queries, not two hundred.
• One location code, and Iranian results differ by city.
• Four days, which says nothing about seasonal movement or the effect of a core update.
• Queries drawn from our own markets; a different industry may behave differently.

What it does establish is narrower and more useful: absent a specific event, the difference between yesterday's position and today's is usually not a difference at all.

You do not need a tool to repeat this. Once a week, at the same hour, record where your ten or twenty main phrases sit and keep the date. Three months later you own a history no product could have sold you. Our manual ways to read your own position are enough to start.

The point where automation earns its keep is when the keyword list outgrows your patience, and recording keyword positions on a schedule is worth paying for then. Not because it shows you the number faster, but because the history gets built even in the weeks you forget. If you only want today's answer for one phrase, a single position lookup covers it.

Last warning, and it has nothing to do with rank. When the position is steady and the clicks are not, the cause is usually elsewhere: the title and description, the load time, or a competitor who simply wrote a better result. Telling a real drop from an apparent one is the right place to begin, and where the build itself is the constraint, a site put together properly from the start makes every one of these measurements easier to act on.

Both snapshots were taken with the same tool and the same parameters. First reading 15-19 August 2026 depending on the query, second reading 19 August 2026. The raw data is ours and we kept it.