02.One story, many URLs
Four results can be one story.
A wire piece picked up by nine sites is nine pieces of coverage and one story. Counting ten is wrong. Counting one is also wrong. Cairn reports both figures and keeps the link between them.
The same page is the same page
URLs are normalised before any comparison: scheme and host lowercased, tracking parameters stripped, fragment dropped, trailing slash removed. Two addresses that differ only by a campaign tag are one address.
One signal never merges two stories
A decision that suppresses a piece as a duplicate needs at least two corroborating signals. A headline of fewer than four content words cannot drive it on its own, because short headlines collide by accident.
The original is the earliest evidenced
Among pieces telling the same story, the one with the earlier evidenced publication date is the original. The others are recorded as pickups of it, and each names the publication it reproduces.
An unknown date cannot win
A piece with no evidenced date sorts last. It can never become the original by default, which is the failure mode of every system that treats a missing value as a small one.
Time windows are part of the rule
Two URLs for one story on one site appear within hours of each other. A pickup is allowed seven days. Beyond that window, two pieces stay two pieces, whatever they look like.
A pickup is coverage
Cairn does not delete syndication. Removing it would understate what actually ran. It is counted, marked as a pickup, and attached to the piece it reproduces, so you can report four pieces and one story and defend both.