Creating a cluster

Three ways to fill a cluster. They can be mixed - a cluster built from a search can be combined with one uploaded from a file, and the result saved as a third.

From search results

Run a search, then use the Cluster button above the results. The domain names are saved, deduplicated, as a new cluster. Where a search returned several pages of one site, the cluster holds the site once.

This is the usual starting point, and the reason to do it even when you do not need a cluster yet: the index moves on, and a search cannot be re-run against the index as it was last month. A cluster is the only record of what a query returned at the time it was run.

From a file

Upload a list of domains, urls or email addresses. Each line is reduced to its domain, and the ones PublicWWW knows about are kept - anything not in the index cannot be browsed or extracted, so it is dropped rather than stored as a name with nothing behind it.

Useful for the other direction: rather than starting from a query, start from your own list of customers, competitors or leads, and use the cluster to filter search results down to it.

From other clusters

Boolean operations build a cluster out of two existing ones:

Subtraction across time is what makes clusters more than a saved list. Store a query's results today, store them again after the next PublicWWW update, and subtract in either direction: one way gives the sites that adopted the signature, the other the sites that lost it. Worked through step by step in tracking a widget change.

Reading a cluster back

Any cluster can be browsed page by page in the interface, downloaded as a file, used as the input to another operation, or extracted for content found on its sites.