McIntosh is the best-documented American maker in this archive, and the most demanding to read. The archive holds 65 model series and 86 serial ranges for the brand, compiled from the McIntosh production history 1949–1999 (Terry DeWick, working from McIntosh Labs data), cross-checked against Roger Russell of McIntosh engineering and the official legacy-product pages. The tables are real. The trap is that they are per model.
The number only works with the model
A McIntosh serial number is not a maker-wide clock. The same five digits live in different years depending on which unit they were stamped on, so the number and the model have to be read together:
- C 22 stereo preamplifier — 1001–7371, 1963–1972.
- C 20 stereo preamplifier: the early run 1001–5817, 1959–1961; the mid run 6001–8379, 1960–1961; the late run 1001–3292, 1961–1963.
- A116 power amplifier — 5000–8515, 1953–1954.
That is why the decoder asks for the model first. Enter a number on its own and the archive will not pretend: it reports the bracket the reading falls into and tells you the model is what narrows it. Type 500D5 without a model and you get 1953–1991 — honest, and nearly useless. Add the model and the table does the work.
Run letters are not years
Numbers such as 500D5 carry letters inside them, and those letters are the second most common misreading after treating the digits as a year. The archive holds the documented run shapes as a system in their own right, so a number can be matched against them instead of guessed at.
When the number and the chassis disagree
McIntosh built in batches from pre-made sub-assemblies. A serial number of one era can sit on a chassis of the next, which is why two documented readings of the same unit are normal here rather than a mistake. The archive's rule is to report the range and leave both readings standing instead of "correcting" either — the answer is a range, not a tidy single year.
What the archive will not do
It will not date a McIntosh from a number alone, and it will not average two sources that disagree. Where a model has no documented range, the page says the model is not covered rather than borrowing a neighbouring year. Values are published separately, each with a basis: the MAC1700, for example, carries a sold-listing reading of $251–$610 from 19 completed listings, with the window stated on the card.
Start from the McIntosh archive — coverage by series, the code systems, and the decoder.