How Query Ewe counts searches
The short version
Each stack is ordered by how often people in the United States searched each term on Google during last full week (Sunday to Saturday). The search numbers on the reveal are estimates, shown with “≈” and rounded to two significant figures.
The order comes straight from Google Trends. The size of the numbers comes from Google Ads search volume. Google doesn’t publish exact weekly search counts for anyone, so every site that shows “searches per week” is estimating. This page explains exactly how we do it.
Query Ewe is not affiliated with Google. We get the data through DataForSEO, a search data provider.
Where the data comes from
| Source | What it gives us | What it doesn’t |
|---|---|---|
| Google Trends (US, web search) | How interest in up to five terms compares, week by week, for the last 12 months. Scores run 0-100. | Real counts. 100 just means “the busiest week for the busiest term in this comparison.” |
| Google Ads search volume (US, English) | The average number of searches per month for each term over the past 12 months. | Anything about this week. Google also rounds these figures into buckets (for example 110K, 135K, 165K). |
How a stack is built
For each stack of five terms:
- Get the weekly trend. Ask Google Trends to compare the five terms over the last 12 months, by week. All five go in the same request, so their scores are on the same scale. Drop the current week if it hasn’t finished.
- Set the order. The order of the last full week’s scores is the answer.
- Find a normal week. Add up the five terms’ average monthly searches from Google Ads, then convert to a week: × 7 ÷ 30.4.
- Find the value of one Trends point. Add up each term’s average Trends score across the 12 months. Divide the normal week (step 3) by that total. The result is “searches per Trends point” for this stack.
- Estimate last week. Multiply each term’s score for last week by searches per point. Round to two significant figures.
Because step 4 compares last week against the term’s own year, a seasonal spike shows up in the numbers. If candy searches double around Halloween, the estimates double too.
Worked example: candy, Halloween week
Data pulled 2 November 2026 for the week of 25-31 October 2026.
| Term | Ads avg monthly searches | Trends, 12-month avg | Trends, last week | Estimated searches last week |
|---|---|---|---|---|
| Skittles | 135,000 | 26.9 | 47 | ≈ 70K |
| Snickers | 110,000 | 24.0 | 42 | ≈ 62K |
| Reese’s | 135,000 | 21.9 | 42 | ≈ 62K |
| M&M’s | 165,000 | 18.5 | 27 | ≈ 40K |
| Kit Kat | 135,000 | 14.2 | 20 | ≈ 30K |
| Stack total | 680,000 | 105.4 | 178 | ≈ 264K |
- Normal week: 680,000 × 7 ÷ 30.4 = 156,579 searches.
- Searches per Trends point: 156,579 ÷ 105.4 = 1,485.
- Skittles last week: 47 × 1,485 = 69,806, shown as ≈ 70K.
- The stack’s total for the week (≈ 264K) is 1.7× a normal week, which is the Halloween spike.
This stack would not be published: Snickers and Reese’s tied at 42 (see the rules below).
Rules every stack must pass
- No near-ties. Neighbouring terms must differ by at least 10% in last week’s score. Otherwise that position is a coin flip.
- Big enough to measure. Every term must score at least 15 last week, and the biggest can be at most 20× the smallest. Trends scores are whole numbers, so a small score is coarse: a term scoring 3 could really be anywhere from 2.5 to 3.5, which is enough to flip its place. This also covers spikes: if one term had a huge week at another time of year (pizza did in June), that week becomes the 100 everything is measured against, the other scores shrink, and the stack fails this rule. A spike in the week we’re asking about is fine: Halloween candy can spike at Halloween.
- Exact wording. The term shown on the card is the text that was measured. We avoid words where Google Ads lumps different spellings together.
- United States, English. All data covers searches in the US.
- Frozen at publish. The data is pulled once when the puzzle is built and never changes after it goes live, even if Google later revises its figures.
Scoring
How points and the share-card squares work is on its own page: Scoring.
Known limits
- The estimates are estimates. Google Ads volumes are rounded, and Trends scores are whole numbers. Expect the counts to be off by 10-20% either way. The order is more reliable than the size.
- Ads and Trends cover slightly different years. Ads uses the last 12 complete months, and Trends uses the last 52 weeks. The gap is a few weeks.
- The two sources count a term differently. Google Trends counts every search that contains the word, so “best pizza near me” adds to “pizza”. Google Ads volume counts the term itself plus close variants such as plurals. The order follows Trends (the broad count); Ads only sets the overall size.