Quarra
From our notes · October 6, 2026

Google Trends is not one dataset

Ask Google Trends the same question from different places and you can get different answers. Not noise that changes every time: in our measurement, a handful of fixed versions, each served to its own share of connections.

What we did

We asked for one curve, “coffee”, United States, past 12 months, weekly, from 43 different US connections, between 16:03 and 16:05 UTC on October 6, 2026.

We grouped the curves that never differ by more than 2 points from each other. That gives 3 versions. The majority version came back 34 times out of 43. The farthest variant is up to 9 points from it: 90 against 99 for the week of May 17, 2026.

The same connections asked for three other everyday terms in the same run. Every term came back in several versions, and 35 of the 42 connections that answered all four stayed on the same side, majority or not, for all four: the version mostly depends on the connection, not on the moment. The Trends website itself does the same; it shows whichever version your connection gets.

60708090100Apr 5May 3Jun 7Jul 599 on others90 on most answers60708090100Apr 5May 3Jun 7Jul 599 on others90 on most answers
Majority versionMinority versionCoffee, US, weekly. Largest gap: 9 points, week of May 17, 2026.
“coffee”, by version
CurveAnswersShare
Majority version3479.1%
Variant 1716.3%
Variant 224.7%
Four everyday terms, same connections, same run
TermVersionsMajorityLargest gap
coffee379.1%9 points
pizza295.3%5 points
weather379.1%5 points
running shoes381.0%3 points

Why it matters

One request, one version

A scraper that asks once hands you whichever version it got. About one time in five, that is not the version most people see.

Comparisons drift

Run the same query from two different connections and the curve can move, with no change in the world behind it.

A succeeded status is not a complete result

Over the 30 days to October 6, 2026, 59.2% of the 22,136 public runs of the most used Trends scraper on the Apify Store ended as succeeded and 22.4% timed out (Apify public statistics).

What our scraper does

It asks each request several times independently, keeps the version most answers carry, and writes the vote on every row, for example sampleAgreement: "5/5": five answers, all five on this version.