Query shapes, branded split and long-tail tiers

These methods sort the search queries of a Google Search Console property into groups. Each method sums clicks and impressions per group. MeasureScale can group queries in five ways:

• by shape: comparison, question, conversational or keyword
• by shape, for each week or month
• as branded or non-branded
• by number of words
• by average ranking position

How to read the answer

• Each answer splits your search traffic into groups and shows the clicks, impressions and share of each group.
• Check how much of your traffic the groups cover first. A low share means the groups describe only part of it.
• A question-shaped query is not proof that an AI assistant served it. Treat the shapes as a proxy.
• The branded split uses your domain name as the brand by default. Give your brand terms if that is wrong, for example on a subdomain.
• For trends, compare the shares of partial weeks or months, not their totals.

What it answers

• How much search traffic comes from question-style or conversational queries?
• Is the share of those queries growing week by week?
• How much traffic depends on people who already know the brand?
• How do short queries perform compared with long ones?
• How many queries rank in the top 3, on page 1 or further down?

Inputs

Every method reads Search Console data for the requested window. With no dates, the window is the last 28 settled days. A settled day is a day whose data Google has finished processing.

• Each query has clicks, impressions and an average position. That is Search Console’s impression-weighted average ranking position.
• Search Console hides some rare queries for privacy. These anonymized queries have no query text, so no method can sort them.
• The two query-shape methods stop at the last settled day. A day that is not settled holds a partial count. The result states the last day they read.

Counting words

Several groups depend on the number of words in a query. A word is a run of letters or digits. An apostrophe inside a word does not split it. A hyphen does.

Query shapes

A query shape describes how a query is phrased. Each query gets one shape only.

A comparison phrased as a question counts as a comparison.

The rules have a version. Both query-shape results state it. The current version is 2026-10-07.1.

Shapes over time

The trend applies the same rules to each period. A period is a week by default. A week is an ISO week that starts on Monday. You can ask for calendar months instead.

The first and last periods can be partial. The result marks them and gives their number of days. Compare the shares of partial periods, not their totals.

Branded split

A branded query names your brand. MeasureScale sorts each query into branded or non-branded. A query is branded when either test passes:

• It contains one of your brand terms. The match ignores case and can fall anywhere in the query.
• It matches your brand pattern, a regular expression. The match ignores case and can fall anywhere in the query.

You can give up to 50 brand terms, each 1 to 100 characters. With no terms and no pattern, the brand term is the domain stem. That is the first part of the host, after a leading “www.”.

Rules for the regex

A badly written regular expression can take very long to run. MeasureScale therefore accepts only a bounded form. It rejects a brand pattern when any of these holds:

• It is longer than 200 characters.
• It has a group followed by a quantifier, such as (...)+, (...)*, (...)? or (...){2}.
• It has more than 3 quantifiers. The quantifiers are +, *, ? and {n}.

Use alternation and quantified characters instead. For example, monster\s?insights|exact[- ]?metrics is accepted. An optional group such as (?:acme)? is rejected, even though it is harmless.

Long-tail tiers

A long-tail query is a long, specific query. MeasureScale sorts each query into a tier by its number of words.

A query with no words is skipped. The result counts these queries.

Position distribution

MeasureScale sorts each query into a bucket by its average position. The boundaries are half-way between whole positions, so each bucket holds the positions that round to its whole numbers. A query at position 3.4 belongs to 1-3. A query at 3.5 belongs to 4-10.

A query with 0 impressions has no position and is skipped. The result also shows the site’s own click-through rate at each position. The CTR curve explains how MeasureScale fits it.

How much traffic the groups cover

The pulls hold only the top queries. Query shapes and the branded split therefore compare the sorted queries with the property’s totals. They report the share of the property’s clicks and impressions that the groups cover. A low share means the groups describe only part of the traffic.

Limits

• A shape is a proxy. A question-shaped query is not proof that an AI assistant or AI Overview served it.
• English rules. The question words and comparison words are English. A question in another language counts as a question only when it ends in “?”.
• Loose comparison matches. Any whole-word “or” makes a 3-word query a comparison. “sooner or later” is a comparison.
• Substring brand terms. A term can match inside a longer word. The term “acme” also matches “acmeville”.
• Subdomain properties. The default stem of https://blog.example.com/ is “blog”. Give your brand terms for such a property.
• Top queries only. The branded split reads only the top 100 queries. The other methods read the top 1,000. Check the coverage before you read a share as site-wide.
• The long tail is thin. The pulls rank queries by clicks. Long queries earn few clicks each, so the top 1,000 hold fewer of them than the site has.
• Search Console row limits. Search Console returns at most 50,000 rows per day and hides anonymized queries. Both remove rare queries first.
• Averages over the window. A query’s position is its average over the window. A query that moved during the window falls in one bucket only.
• Full pulls. Every result says when its pull was full.