Not being a computer genius, I offer this layperson’s interpretation:
Every variable your search contains—every date in a range, every place of residence, every keyword—computationally is a separate query that runs through the millions of records in Ancestry.com’s servers.
The search engine operates on an algorithm that assigns each record points based on terms in your search that match data fields in the record. Some data fields, such as the name, are weighted more heavily than others (that is, a matching name would get more points than a matching place of origin).
The search engine also assumes some terms are the same, for example, Kathleen and Cathy (who knew there are 800 variations on the name Catherine?), Florida and Fla, Syria and Alssyria. And it tries to account for the variations in spellings, the roaming birth dates and other unexpected information in historical records. Search product manager Anne Mitchell calls this “fuzziness.”
That’s why some records in your search results seem far outside the realm of possibility for your ancestor—the date or place may have been off, but the other stuff was close enough to get the points necessary to make the list.
Frustratingly, sometimes records you know aren’t your ancestor get more points than the ones that might be him. You could spend hours sifting through all the search results—it’s hard to know when to stop (someone said after two or three pages of results, it’s unlikely you’ll find the record you’re looking for).
Mitchell said that the search engine’s algorithm will soon be adjusted to subtract points when a name or date in a record doesn’t match what you typed in. Before, this additional step in the search process would’ve taken too long and made the servers start smoking. But now that the engineers have almost figured it out, your search results should appear in a more logical order, with the best matches higher up on the list.
It’s entirely possible my ancestors’ passenger list has been destroyed and they hid from the 1920 census enumerators, but once the changes go live, I’m going to repeat these frustrating searches.
Something else to think about if you have an Ancestry Family Tree: Family trees product manager Kenny Freestone said the quality of a family tree search—the automated search that give you those “shaky leaf” hints next to individuals in your tree—is more precise than for a ranked search. That’s because the hints are based on several generations of your tree, rather than just one person.
(And, by the way, you now can hide a tree so it’s completely excluded from the index.)