When researching the 1600 VICmodem, I found an assertion that the 1600 was the first modem to sell a million units. And I’ve repeated that claim myself at least once. But I never felt quite right about that. I am a classically trained journalist, and if you have the ability to double check a claim, you are supposed to do that. In this blog post, I try to get a good estimate of how many VICmodems Commodore sold.
Answering the question with one shot
I had a chance to possibly settle the million-modem argument with one shot. I had a VICmodem. We bought it in 1986 from a liquidator. So it had a chance to be among the last units produced. All I needed to do was find it and look at the serial number. If it had a 7-digit serial number, it might instantly resolve the question.
Peeking out of the top of a very ordinary moving box, I caught a glimpse of the original box the modem came in. It was buried, but reachable.
I retrieved the box, and, with much anticipation, opened it. Empty. Discouraged, I shined my flashlight into the moving box and hoped for the best. The modem was loose inside it, next to an old video cable. I grabbed the modem, flipped it over, shined my flashlight onto it to get a really good look, and read the number.
37225.
Well that was a disappointment. It was much lower than I expected for 1986 production. And it sure wasn’t high enough to prove much of anything. I’d need an alternative method.
Using the German Tank Problem to estimate the number
So I used the German tank problem methodology to try to estimate the number of 1600 modems Commodore sold. If this name sounds familiar, it’s the methodology that the Allies used to figure out how quickly the Germans were capable of producing tanks during World War II. If the estimation method was good enough to win a war, it’s good enough to estimate sales of retro computer gear when no official figures are available. It’s also been previously used to estimate the number of 1581 disk drives sold.
No lives are at stake in this case. We just want to know if a million of these modems sold. And if the number was less than a million, we’d like to arrive at a reasonable estimate of the actual count.
I started by looking at available Ebay listings. The first few serial numbers I observed were four or five digits. They seemed low. Not at all what I expected from a million-selling product, but not enough to prove much either. So I went and gathered as many VICmodem serial numbers as I could. I checked Ebay listings, photos on social media, Youtube mail call videos, and anywhere else I could think of.
Over the course of several years, I accumulated 61 unique serial numbers, plus my own. There are always a few of them on eBay. It seems that people are reluctant to throw them out, but there also isn’t a ton of demand for them. They aren’t something you will use for anything now, so if you already have one, you don’t really need a second one. The ones that seem to sell either have unusually low serial numbers, a price under $10, or are complete in box.
I also saw a fair number of the same serial numbers over and over again. Some of these have been for sale a very long time.
What the VICmodem serial numbers looked like
I observed six serial numbers with four-digit counts, so while it is possible Commodore may have skipped some ranges by not starting at serial number one, it doesn’t look like it will affect the total count by much. The lowest serial number I observed was 1281.
I only observed seven six-digit serial numbers. The highest legible number was 122,048. I did observe one that I could tell was six digits, but only the first two digits were legible due to damage on the sticker.
That left 48 serial numbers in the five digit range. One of those was damaged, but I could tell it was 5 digits, started with a 4, and looked like it could be 44909. When 48 of the 61 observed samples are below 100,000, that seems like a strong suggestion the number of units sold is closer to 100,000 than 1 million, before I even do any math.
Hours after posting this, Bluesky user Ramblingfan pointed out serial number 135398 is in the Smithsonian’s collection. So that plus a few more serial numbers I received pushed the overall estimate up a bit.
So, how many model 1600 VICmodems did Commodore sell?
Based on the methodology, my initial estimate was 124,307. If Commodore started at serial number 1000 instead of serial number 1, that knocks it down to 123,307.
I hate to say it, but there’s no way Commodore sold a million VICmodems. If they had, 6-digit numbers would be much more common than 5-digit numbers. And they aren’t.
After I got a few more serial numbers, including 135398, that pushed the estimate up to 137,581. Or, if Commodore started at serial number 1000, 136,581 units.
Here’s the math, if you’re curious:
Population Maximum = 135398 + (135398 / 61) – 1 = 137,581
135398 is the highest serial number I observed, and I observed a total of 61 serial numbers.
As for how I ended up acquiring serial number 37225 when I bought it in 1986, it may have been a refurbished unit. Or maybe it was just always on the bottom of the stack in the warehouse.
Dealing with the margin of error
The German tank methodology has its own margin of error built in. You divide the largest serial number you observed by the number of samples you collected. The larger the number, the smaller the margin of error. With 61 examples observed so far, my margin of error is 2,183.
The German tank problem is designed to estimate high. But when I apply conventional statistics to its estimate, with the small sample size, I can be 70% confident the actual number produced is within 7 percent of the estimate. That would put the upper bound at 147,212 units.
Having more serial numbers increases the confidence in the count. But even with the small sample size, the margin of error isn’t high enough to close that gap to a million. I’d feel much better about the overall number if I had 140 serial numbers. With that high of a number, we could estimate the overall count to within 1,000 units.
Estimating with anecdotal evidence
Anecdotal evidence is much less reliable. But for what little it’s worth, 120,000-135,000 units is in line with my experience. We can be pretty confident that Commodore sold 12 million Commodore 64s and 128s because they disclosed counts in their annual reports. There is no data for 1994, but we know that number wouldn’t have been enormous because 1993 sales were extremely low, and Commodore only survived one quarter out of four in 1994. We are missing one other year from the 1980s but we have figures for the years around it, so the average of those two years is a reasonable estimate. Commodore also sold around 2 million VIC-20s. So that brings the audience to 14 million.
If the VICmodem had sold a million units, that would mean approximately 10% of Commodore owners bought one. Yes, I know a million isn’t 10% of 14 million, but the VICmodem was discontinued in 1986. The Commodore 64 still had 2/3 of its life left at that point.
I knew a lot of people in the 1980s with Commodore computers, possibly 100 or more. So I would expect to have known nine other people besides me who had one. I only knew one.
Even though VICmodem sales almost certainly fell well short of a million units, 135,000 units by 1986 is still a lot of modems for that time. It wouldn’t surprise me at all if it was the first modem to sell 100,000 units. And maybe that’s what everyone involved is remembering, and they just accidentally added a zero.
David Farquhar is a computer security professional, entrepreneur, and author. He has written professionally about computers since 1991, so he was writing about retro computers when they were still new. He has been working in IT professionally since 1994 and has specialized in vulnerability management since 2013. He holds Security+ and CISSP certifications. Today he blogs five times a week, mostly about retro computers and retro gaming covering the time period from 1975 to 2000.