What building a rank tracker teaches you about Google
Six months of parsing Google results at scale: what building a rank tracker teaches you about SERPs that reading about them never will.
We spend our days fetching and parsing Google results pages — it’s the unglamorous core of building a rank tracker. Do that at volume for a while and you develop opinions that reading about SEO never quite delivers. Some field notes from the parsing mines.
SERPs disagree with themselves, constantly
Fetch the same query twice, minutes apart, held to the same country and device, and a noticeable share of the time you get different pages — an order swapped here, a result rotated out there. Not because anything changed in the world; because Google is a distributed system running live experiments, and “the results for query X” is a statistical object, not a fact. Every rank tracker quietly makes peace with this; now you know why your rankings wobble even when your tracker is doing everything right.
”Position” is a parsing decision
Where exactly is position 3 on a page containing two ads, a map pack, a video carousel, and a “People also ask” box that expands into infinity? Every tracker answers slightly differently — which features count as slots, which get skipped, how a two-URL sitelink block is numbered. The convention we use (count organic results, skip the furniture) is defensible and common, but it is a convention. If two tools report different positions for the same keyword, this is usually why — neither is lying; they’re measuring with different rulers.
Layout churn dwarfs algorithm churn
The famous updates get headlines, but week to week, what changes most is the page itself: features appearing for query classes that never had them, layouts tested and reverted, snippet formats mutating. Parsing code that was perfect in October is quietly wrong by March. For site owners the implication is practical: when a keyword’s clicks change but its position didn’t, suspect the scenery first.
Countries are different planets
The same query, one border apart, produces different results and often different kinds of results — more shopping features here, an answer box there, entirely different local packs. Anyone selling you “your global ranking” as one number is averaging planets. It’s why we track per country, per device, and why the combination — not the keyword — is the real unit of measurement.
The meta-lesson
Building the measuring instrument makes you humble about the measurements. Rankings are real and worth tracking — trends, especially, are honest — but a single position on a single day is a photograph of moving water. Design your decisions (and your reports) around the current, not the splash.