FEAL
NTT's fast DES alternative from 1987, remembered mainly as the punching bag on which nearly every new cryptanalytic technique got its first workout.
Akihiro Shimizu and Shoji Miyaguchi wanted a Feistel cipher that ran quickly in software on the 8-bit processors of the 1980s, so they dropped DES-style table lookups in favour of addition and rotation. The original had four rounds. Bert den Boer broke it in 1988, and the response was FEAL-8, FEAL-N and FEAL-NX, with more rounds and longer keys.
Every version fell. Biham and Shamir used FEAL as a showcase when they presented differential cryptanalysis, and the work by Matsui that led to linear cryptanalysis started there too — FEAL-4 yields to a handful of known plaintexts. The cipher never caught on in practice, but as teaching material on how not to do it, it has few rivals.