Cannibalization

Keyword cannibalization happens when several pages of one site compete for the same search query. Google then shows different pages for that query, and no single page collects all of its impressions. MeasureScale uses this method to find those queries and to measure how evenly each one splits.

How to read the answer

• Each result is a query that sends impressions to two or more of your pages.
• A severe result means the pages split the impressions almost evenly. A low result means one page holds most of them.
• Start with the queries at the top of the list. They combine a large demand with an uneven split.
• A split does not prove you lose traffic. It shows only that impressions are spread.
• A page that replaced another during the window also shows as a split. Ask your assistant to check the dates of recent page changes.

What it answers

• Which search queries send impressions to two or more of your pages?
• How evenly does each query split between those pages?
• Which conflicts involve the most search demand?

Inputs

MeasureScale reads Google 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.

• The pull is the top 5,000 query and page pairs by clicks.
• MeasureScale sums clicks and impressions for each query and page. It weights position by impressions.
• It excludes anonymized queries. Search Console hides some rare queries for privacy and reports their traffic with no query text.
• It ignores a page that has 0 impressions for a query.

How it works

For each query, MeasureScale works out each page’s share of the query’s impressions. It then scores how evenly those shares spread.

• The normalized conflict index (NCI) is that score. It is always between 0 and 1.
• NCI is 1 for a perfectly even split.
• NCI is close to 0 when one page holds almost all impressions.
• A raw spread score is also reported. It grows with the number of pages, so it can mislead.
• Severity and the conflict filter therefore use NCI, never the raw score.

Severity

Filters and order

A query is listed only when all of these conditions hold:

• At least 2 pages have impressions for it.
• Its total impressions reach the minimum. The default is 100. You can ask for a smaller or larger minimum.
• Its NCI reaches the minimum. The default is 0, so every split is listed. You can ask for a higher minimum.

The list ranks queries by both total impressions and NCI, highest first. Ties are sorted by query. A large query with a moderate split can rank above a small query with an even split.

Limits

• Row cap. The pull stops at 5,000 query and page pairs. Long-tail queries and pages outside that set are missing. The result says when the pull reached the cap.
• Search Console limits. Search Console itself exposes at most 50,000 rows per day for a property. This limit can also hide long-tail pages.
• Impressions, not intent. NCI shows that impressions split. It does not show whether the split costs traffic.
• One window. MeasureScale sums the whole window. A page that replaced another page during the window shows as a split.
• Anonymized queries. Queries that Search Console hides are not analyzed.