HP serial numbers

[PeterG]

On HP 15Cs made in the USA serial numbers appear on the top edge in the form yywwAnnnnn where the number yy + 60 denotes the year; ww denotes the week produced; A, the USA (B for Brasil) and nnnnn the sequential number of the calculator produced in that particular week.

Consider the following serial numbers of HP 15C calculators:

2505A79855 on Ebay
2638A19369 on Ebay
2707A54723 on Wikipedia
2350B38564 on Ebay
And my two:
2249A0994 and 2413A45059
And from Brasil 2350B38564, on Ebay.

Although the sample is small we can deduce from calculator number 79,855 that at least 80,000 were produced a week in the USA in 1985. 80,000 a week for 50 weeks a year for 7 years (1982 to 1989) is 28 million with an additional 14 million or more from Brazil (assuming 40,000 per week given the Brazilian serial number 38,564) giving an estimated total of 42 million.

This seems to me to be an implausible number even without considering the other models that HP was producing at the same time. Could it be that the nnnnn component of the 15C serial number is not the number produced in a week but rather the sequential number produced in the year?

[rprosperi]

All evidence, and reports from former employees, have confirmed these are weekly production numbers. As for running annual totals, your unit 2249A0994 would imply that they only made 994 units by Dec 1982, so clearly not the case.

I think the fallacy of your math is assuming they are manufacturing continuously, every week of the year; it doesn't work that way. They would manufacture xx,000 at a go, possibly every 6 weeks, or monthly, or some other cycle, based on sales forecasts. There's no reason to make (and then ship, receive, warehouse, etc.) them on a weekly cycle, it's far more cost effective to make them in batches.

If the sales forecast was e.g. 5k units / week, they surely would not manufacture 5k, then ship them, etc. in such small lots, they would make 10-20k units, over some number of days/weeks and then ship the lot.

[PeterG]

Food for thought; although unit 2249A0994 in no way implies that only 994 units were made. If 80,000 units were made that week 2249A0994 is just one of them; someone could have bought 00001 and someone else 79999.

I have had experience of working in a factory in 1944 (I am 91) and at that time assembly lines had to be kept going for maximum efficiency.

Do you happen to know how many 15Cs were actually made?

[Steve Simpkin]

Tangentially related to the quantity of HP-15C units that HP produced.

From Dr. William Kahan regarding the quantity of HP-15C that he estimated HP could sell per year:

KAHAN: “The HP 12C was successful enough that they were willing to take my advice about building the 15C, but not take my advice about how many to build. They wanted a third of my figure, and Dennis Harms did half again what they wanted, and that’s what they were doing. They were producing half of my number. … The marketing people had a third of my number, and Harms ended up with the production line producing half of my target number, and then these calculators were disappearing off the shelves as fast as they could be supplied. MIT, for example, for a couple of years was telling its freshmen that they should buy the calculator, and it had a special deal with HP that would get them involved in a somewhat lower price.
HAIGH: What year did the calculator appear?
KAHAN: I think it was 1982, give or take. We could probably look at the manuals and find out exactly when it appeared. Well, HP never produced an advertisement for the 15C in its own right in any Western language. It might have had advertisements that listed the 15C and the 16C and the 11C and so on, but never for the 15C in its own right except for one in Japanese. I saw an advertisement in Japanese. They were selling them by word of mouth as fast as they could produce them, and when my friends and I who had worked on this went to the marketing people and tried to persuade them, “Look, set up another production line, because you want to gather your flowers while you may,” they said, “No, if we set up another production line, we may end up with inventory after all. You know how sales go. The sales rise to a peak and then they go down,” and these guys thought that they must be hitting the peak. They weren’t hitting a peak. They were hitting a ceiling. It’s a different thing. So they never did set up another production line. In consequence, the market was starved. There were waiting lists, and that window closed, and so I never did get the calculators into the hands of sufficiently many students to change the ways in which professors would issue assignments, and that was a bitter disappointment. It colored my relations with this particular group at HP. I continued to work with them for a couple of years, but my heart just wasn’t in it anymore.“

More about the HP-15C, how it was manufactured and other models that Dr Kahan worked on at the following oral history.

http://history.siam.org/pdfs2/Kahan_final.pdf

[PeterG]

Thank you Bob and Steve. I do now see that Bob is correct and that each 15C has a unique serial number for a particular week. Unfortunately it is no guide to how many were produced in a week.

I have read through Professor Kahan fascinating although somewhat periphrastic interview, linked by Steve, and here is a further extract relevant to 15C production:

KAHAN: “Well, when we then had these things going, it was time to build the HP 15C, and I had to persuade them to do it. Here, again, is an unfortunate story. …… I came to some conclusions about how many should be produced, and that was part of my argument for producing this calculator. The people with whom I worked most directly also wanted to produce this calculator. The marketing people, however, had done their own survey; they came out with some number which was half of mine, and I have no idea how they got it. It could be that they just took my number and said, “Oh, he’s an enthusiast. Let’s just simply half it.” Maybe they did that, or maybe they did their own research. I don’t know what they did. This was to have a devastating impact later because, after we got the calculator microcoded and everything was done and we had it tested and running, the marketing people then told one of our guys how many they wanted produced. It was this half of the number I wanted.
Now, fortunately, the guy they told it to was Dennis Harms, and what Harms did was to build a roboticized production line in a room that was about the floor plan of this house, maybe a bit smaller. He had a roboticized assembly line which, on the end, cost Hewlett-Packard 20 seconds of human time per calculator, from unloading the stuff they purchased from vendors at the loading dock to loading the cartons full of crated computers to be shipped out to their sales outlets. Twenty seconds of human time per calculator. This included jigs that tested partially assembled components. Now, if my memory serves me rightly, this is one of the tricks they used: a partially assembled subcomponent went into the jig; it got tested. If the test failed, this component got moved by the robot and put into a bin of rejects. Otherwise, it went to the next stage on the assembly line. And these things went faster than fingers. I tell you, I saw it working. They really went fast.
What happened to the bin of rejects? Well, somebody went through the bin of rejects and looked to see whether the rejected parts had a little red dot of paint on them and, if so, they went to the crusher. Otherwise, they get a little red dot of paint, and they go back into the supply bins. Of course, it was this partial assembly that went into the bin, but it was then disassembled; the individual components on the assembly got these dots, and they went into the stacks that were going to be used. So, you see, instead of having to diagnose an assembly to see which of its parts was defective, what they did was try it again later, and if it got rejected twice, then forget about it. All right, so some good parts would get rejected, and on very, very rare occasions, perhaps, something defective might get through. But they had very few returns. Texas Instruments used to have a return rate that ranged from a third to a half of their calculators being returned as defective. I don’t think Hewlett-Packard’s return rate for all causes, including the fact that somebody regretted the purchase, ever got as high as ten percent. So these calculators were being produced and, my guess is, it cost them something under $15 to produce, maybe, as little as 12 and you send it off short, but it gets sold in CompUSA for 70 bucks. So there’s a lot of room for mark-up along the way.”

Quite astonishing that it took about 20 seconds of human time to produce a 15C calculator.

[BruceH]

(03-13-2022 06:44 PM)PeterG Wrote:  I have had experience of working in a factory in 1944 (I am 91) and at that time assembly lines had to be kept going for maximum efficiency.

That's very true and the 15C assembly line may well have been kept busy by switching it over to making batches of 12Cs or 11Cs.

The test would be to compare serial numbers of 10s, 11s, 12s and 15s: if the above were the case then you wouldn't expect to ever see two different models manufactured in the same week.

(I'll caveat that by saying that 12Cs were made in such large volumes they probably had their own line.)