The Japanese Typewriter Datecode Theory (a work in progress)

There’s a theory I’ve been pondering – a datecode theory that is exceedingly simple and once you think about it, surprisingly obvious and relatively easy to prove now that we have quite a lot of example machine galleries in the TWDB and someone who has devoted his collection and research to the age of Electronic Typewriters (ETs) and has the knowledge and documentation to shed some light on the darkness and lack of info we have on typewriter manufacture and serial numbers through the 1980’s and 1990’s. That person is Steve Kuterescz, who runs the “ETZone” blog, and he’s the guy I’ve collaborated with to put some meat on the bones of this theory. He’s aces at finding the date ranges that certain ET models were produced and advertised, and it’s that information that confirms that the datecoded serials of machines we have in TWDB galleries and his own research correlate well with the date ranges we would expect to find them.

The first inkling I had of the Japanese Datecode Theory was in 2014, when I finally cracked the code for Brother machines. That one was made easier and more certain because I had actual Brother documentation detailing the datecode method, and lots of supporting documentation. It should be noted that the OMEF was the previous primary source for the earlier age list, but they didn’t have it quite right, or at least they didn’t explain it sufficiently. We’ll get back to the OMEF.

Chinese-Made Royals
Not long after that, I was working on the Royal page, trying to include all of the Royal branded machines that were made by other manufacturers. Of course, towards the end of the page, that meant including the Chinese-made Royal-branded Japanese-design machines made in the 2000’s through today. When examining them, I discovered another, very obvious and easy to read datecoding in the serial numbers on these machines. I noted that as a clue that datecoding serials might be something other manufacturers than Brother did.

Then, about a year ago, I saw a pattern in Silver-Seiko serial numbers where they start in 1967 in the 7 million range, and are at the 1 million range in 1971 (by documented purchase date of several machines) and seem to group in batches sometimes years apart. I can certainly see contract work for Royal being very much batches of machines. What I can’t see is Silver Seiko pumping out a million machines a year, so a datecode like Brother’s, except no month code letter, is the logical conclusion. Two flies in the ointment, though – I could find no “leading zero” serial numbers – but there are 10 million series ones – my current thought is that those might be the 1970 machines that I expected to have leading zero serials, but the serials in the 10s of millions range are all non-US machines, so that could also be a region code. The second fly? The OMEF numbers – both for Silver-Reed and for Silver-Seiko made Royals. They don’t follow datecode at all.

In fact, it was the OMEF that kept me from really investing in the Japanese Datecode Theory for *other* Japanese manufacturers until I noted a strange thing in the OMEF’s Facit age list that suggested that Facit had switched to a datecoded serial number around 1975, and the OMEF numbers were occasionally nonsensical afterwards. It seemed clear that there was a point where OMEF stopped getting manufacturer numbers and started getting numbers sourced way farther down the distribution chain. That doubt made me suspicious of what the OMEF says about Silver-Seikos as well.

Then I see this machine:
Yeah, serial number 7999307, which I put in 1967, but which would theoretically mean that Silver Seiko *did* make almost a million machines a year. I take a step back and look at the BMYS page for the rest of the Royal Mercuries:
Oh hey, would you look at that! When properly lined up by serial number, we see that we have no Mercury in 1967 that has a series number lower than 400k. In fact, we have every “hundred thousands” position accounted for from 4 through 9, except below 400k. “Self”, I says to myself, “This could mean one of two things.” The first; maybe they started at 999,999 and counted down? That’s belied immediately in 1968, when we see the series start in the 100k range. So not that. What’s left? Maybe the second digit in the serial number stands for a grouping of weeks, and production started in the 4th grouping of 1967? It can’t be a month counter or week counter – you’d need 2 digits for that. Is there any way to divide a year into 10 (0-9) parts? 1968 accounts for serials in the 1,2,4 and 500k groupings, which suggests that Silver-Seiko spent 10 of these time periods building out the first batch of Royal Mercuries for Royal to sell over several years. Then no more Mercuries (though SS does produce other models for Royal from 1968 through 1971, especially the plastic-shelled models) until 1971, when it appears that Silver Seiko began a batch of Mercuries in the 1967 pattern for at least 9 of these time periods, then spent 4 periods redesigning the ribbon cover to add “Litton” to the logo and add dark brown stickers all over it. Those latter 1972 Mercuries were made for at least 3 periods. We know from ads that this style debuted in 1972 and was sold till at least 1974. Now we know that *all of them* were made in 1972.

But like I said – there’s some doubt there, mainly due to what the OMEF says. If the Datecode theory is correct, then the OMEF is basically trash for Silver-Seiko. What I do know is that dating the machines by this presumed datecode puts machines into age ranges that corresponds well with period advertising and verified sale dates, and in the case of Silver-Seikos built for other brands from Royal to Boots department stores, causes the machines to group into a few distinct years – with a model sometimes split into more than one manufacture run that can be years apart. This is exactly what I’d expect to see with machines ordered in batches and exported. The upshot? I’m pretty sure that Silver Seiko used a datecoded serial number and I’ve re-dated lots of Silver Seikos in the TWDB (this is how I discovered that doing this arranges them in really nice manufacture date batches). I haven’t changed the Silver-Seiko serial number pages yet, though – I feel the need to clarify some of the doubt that swirls around still.

So then we get to Nakajima. I’ve suspected that Nakajima used a “first digit of serial is last digit of year” datecode for awhile, but just didn’t have sufficient corroborating evidence for the idea until I started hitting it sideways by looking at Nakajima-made machines sold by Sears, Olympia, Royal and other OEMs. As an example, Sears purchased Nakajimas in a couple of specific instances that are clearly single batch orders, and we know when they were sold thanks to Sears Catalogs:

3rd iteration “Sears Portable”, sold from 1972 through 1974, datecode suggests they were all made in a single batch in 1971:

5th iteration “Sears Portable”, sold presumably in 1989-90 (tho I don’t have a catalog for that period – I would expect that when found, will be in that catalog) datecode *and* the logo style confirm 1989.

1st iteration “Sears Manual 1”, sold only in 1974, examples have serials starting with “4”.

Comparing Olympia B-12’s and S-10’s through the Nakajima-made electric typebars sold by Olympia, Sears, Facit and Royal also show datecodes that are sensical for the years expected.

I had Steve take a look at his research and check the idea for validity, and he replied with even more examples that confirmed the idea. Have I updated the Nakajima page yet? Not yet – I feel we’re right, but I need to collect a list of expected model year ranges to help out with the decade determination.

In February, I was reading a blog post Steve did on Panasonics, and it was just immediately obvious to me that the serial numbers in the article followed a Brother-style numeric/alphanumeric datecode. I put that idea to Steve to check out using his research, and he easily confirmed the idea. A new Panasonic page was built on TWDB immediately, it was so obvious and checked out so well. Panasonic was yet another stone on the scale suggesting that most, and maybe even *all* Japanese manufacturers used datecoded serial numbers following slightly different patterns.

So then there’s Sharp. This is where the dam really broke for me to the point where I feel I can post about this theory with confidence. I was bothered by Sharp – they look like datecoded serials, but I didn’t have the information to check it, and what I did know made me *doubt* that Sharp used a datecode. So I emailed what I was thinking to Steve and asked him to find me some proof that would *disprove* the theory. I figured that’d be easy to do with his research. Just find me a couple machines with datecodes that can be proved *not* to fit in the date range they’d be expected.

And a couple days later I get an email back telling me that not only could he not disprove it, the evidence he’d collected made him certain that Sharp numbers *were* datecoded. He made a blog post detailing his findings, and last night the new Sharp page went up at TWDB.

So, now we’re at today. Right now Canon is the odd man out in my estimation. I can’t make Canons fit into a datecode given when we currently think they were produced. I’ve put Steve on the case, as I’ve exhausted my leads. All he has to do is show that datecodes *don’t* work for Canon serial numbers, but who knows what the return email will say?

Now here’s a curious one. Facit used their own serial series for most of their history, but the OMEF shows weird, huge series jumps for many models after 1975. Their serial number scheme for manual standards and electric typebars seems to have changed to a Nakajima-style datecode. This isn’t especially surprising to see this, as we know that Facit sold Nakajima-made ETs as well towards the end, and it would have made sense to also switch the last of the other product lines to match that scheme.

It’s Facit’s numbers in the 1980 OMEF that suggest strongly that the OMEF at some point had stopped getting reliable manufacturer data from some OEMs around the mid-70’s and had started getting samplings of serial numbers much farther down the distribution chain, probably they had resorted to polling their member dealers for current stock as you start seeing serial numbers listed that are pretty obvious from the prior year’s manufacture. It’s that sort of thing that makes me more comfortable with the idea that the OMEF probably had no real manufacturer data from Silver Seiko, and got that data from farther downstream, older stock and much less reliable sources. And if we’re right, their numbers for Silver Seiko are absolute trash.

Frankly, with the benefit of lots more information that used to not be available, the idea of datecoded serial numbers being a fairly common thing with Japanese manufacturers seems real obvious. Considering the fact that Japanese industry was basically rebooted and vastly rebuilt in the post-war period, it wouldn’t be especially strange to see them adopting newer, modern ideas about record keeping and it wouldn’t be especially strange if they all adopted similar systems at roughly similar times. If this is so, why did the OMEF not understand that these companies used such a system? It would’ve made their job far easier if they had. Instead, it looks like they just assumed that these OEMs used old-style straight series numbers and it may just not have ever become common knowledge among dealers that Japanese OEMs used such a system. That’s easier to believe than the idea that it *was* common knowledge and everyone just forgot.

Of course, the issue in locking it all down and rebuilding the pages for brands like Silver-Seiko and Nakajima will be in painstakingly gathering the evidence for *each* manufacturer to confirm or disprove the theory for each one. We will be looking at them all and making that sausage over the coming weeks/months. Confidence is very high, though. I think we’re going to be shining light on that whole swath of typewriter production from the late 70’s through the late 90’s/early oughts. That’s a great thing that I wasn’t sure we’d ever be able to do when I started this project a decade ago. Warm fuzzies – yep. Very warm fuzzies (:

Updated: August 14, 2022 — 3:48 pm


Add a Comment
  1. Excellent detective work on your part, Ted. I hadn’t given the serial numbers in my e.t. archive much thought until I was prodded. Happy to give something back … ;)

  2. When trying to figure out the logic of serial numbers, didn’t folks take a look at known info from manufacturers other products, and product enthusiasts, for clues? For example Seiko’s numbering system for their watches is well-known by now:
    Similarly, folks into retro electronics know that Panasonic and other companies use dates in their serial numbers. Might be useful to go a little cross-disciplinary to solve the details of this puzzle.

    1. True, and thanks for that Seiko link – that’s excellent confirmation that a datecode is probably the case, but I’m only seeing a second digit range of 0-9 in Seiko typewriter serials where watches use 1-9+O,N,D. There’s part of the rub – these were different divisions of the company and may not have used exactly the same system. The second part of the rub is that we have a Primary, contemporary industry source that is entirely wrong on serial numbers/dates for this manufacturer (while being reliable on others) and that insists that the serial numbers are a straight series like most other typewriter serial numbering schemes of the time. It can be tough to overcome such a barrier and suss out the truth. (:

      But suss it out we will, and I’ll take any clue that someone wants to point me to, if I’ve missed it. :D

  3. Congrats for this achievement after a colossal research effort!

  4. I just uploaded a gallery of my Nakajima ALL 520. I had thought that it was manufactured in 1989, as I read (on the TWDB) that the company didn’t rebrand to “Nakajima ALL” until 1988-1989, and that models such as the 500 & 530 showed up in 1989. Mine being a Nakajima ALL model 520 convinced me that it had to be made in 1989, based on that information. I noticed that my gallery was edited/updated to say that my machine is a 1987 model. I am assuming it is because the serial number starts with a 7, but then wouldn’t 1987 predate the “Nakajima ALL” branding as well as the “520” model number?

    1. Company histories may not be exactly accurate, but we’re pretty sure the datecodes are. (:

  5. Great analysis! Hooray for the new insights that these crowd-sources databases can deliver – also a prodict of the Internet-age :-)
    (Have been similarly diving into the SIM serials – darn, should share… so many things!)

  6. I’ve got an Apollo 12-GT s/n 2943226 and I was looking at the DB and scratching my head just like you. “Surely there’s no way they made that many typewriters in this year!” So, by what you have posted, I then would expect that this was made just at the end of 1972. Kind of surprised to find that the series was all produced in a single year and something strikes me as odd about that, so I think more research is needed. Too bad we don’t have access to the actual company records.

    Many earlier makers simply serialized their manufacturing or, at most, skipped to a new batch every year/model/revision. (Like 108950 for 12/31 or the end of “revision 1” and the next made bears s/n 200101.) I know that, today, APC uses a year and week dated code in their serial numbers, so such things are used, but who knows back then.

    1. We know because of period ads and product cycle logic. it’s 1972. (:

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.