We study an instance of a general production problem called the 'match
ing problem', which involves extremely tight mating tolerances for ass
embly. Situations exist in industry where the production equipment use
d to produce components cannot meet the tolerances required for assemb
ly. In these situations components are often sorted into different siz
e classes and then 'mated' with other components of the same size clas
s. As a specific instance of the general matching problem, we focus on
the production control of components (needles and bodies) used in fue
l injectors in an actual automobile subassembly production facility. I
n this setting the manufacturer sorts components into class sizes such
that a needle of class i can be matched with a body of class i to sat
isfy the tolerance requirements. We study dynamic control rules that a
ttempt to optimize the operation of this production facility within re
asonable computational limits; we further develop a bound on optimal s
ystem performance. In the application studied, the performance of the
recommended heuristic is within 0.2% of a bound on optimal performance
.