The encyclopedia · Software & IT · Technical decision · 1918–1930
Soundex turned any surname into one letter and three digits
Russell's 1918 patent encoded names by sound, so Smith and Smyth filed together in census indexes.
Robert C. Russell
The solution
Before computers, finding one person in a census meant knowing exactly how their name was spelled — but record keepers wrote Smith, Smyth and Smythe. Robert C. Russell's 1918 patent (US 1,261,167) proposed indexing names by how they sound.
The Soundex code keeps the surname's first letter and assigns three digits to the remaining consonant groups, discarding vowels and repeated letters. Any spelling that sounds the same produces the same code, so related entries are filed together instead of scattered.
The National Archives still documents the rule — every code is a letter plus three numbers such as W-252 for Washington. From the 1930s, the Census Bureau and the WPA built Soundex card indexes for decades of census records, letting clerks find people without knowing the exact spelling.
Why it worked
- Phonetic groups collapse spelling variants into one code
- A fixed four-character key is cheap to file and find
- Clerks needed no spelling knowledge to search
- It scaled to millions of census cards
What can be applied
When data is messy in predictable ways, encode the equivalence class instead of cleaning every record: a lossy key that groups near-duplicates makes lookups robust to error.
Aftermath
The Soundex indexes became a core genealogy tool, and the phonetic-coding idea spread to name-matching systems, voter files and modern record linkage — Russell's 1918 patent is cited as the origin of phonetic name matching.
Sources
spotted an error? The archive wants to know.