Many articles claim that there were 30 billion COBOL transactions per day. It is one of two ingredients of the „200 times more COBOL transactions than Google searches“ claim. After I had thoroughly debunked the other ingredient, I was wondering if the 30 billion COBOL transactions had ever been correct in the first place.
So I started digging into this number also and I have to say that it is really hard to find reliable information on this topic. The numbers become fuzzy in the 1990s for many reasons. The wayback machine started in 1996 and wasn’t as exhaustive back then, so references become unobtainable. Print-Magazines of the time can often be searched on Google booksearch, but the content can often not be examined beyond short excerpts.
I had already given up on it, but then I got creative.
It is a somewhat adventurous feeling. Like Indiana Jones, uncovering hidden tombs, opening ancient graves, looking for further clues to the location of the holy grail.
30 billion "COBOL" transactions
The earliest source for 30 billion „COBOL“ transactions I could find at first is this publication from 2007, which simply attributes the number to „Gartner“. At first I was surprised, not to have found earlier sources, but then I found the computerworld article 35 Technologies that shaped the industry from 2002.
Interestingly, the computerworld article says that there were 30 billion CICS transactions per day – not COBOL transactions! I don’t distinguish between COBOL and CICS much, because 70% of all mainframe programs are written in COBOL anyways, and I doubt that FORTRAN or PL/1 programs involve CICS much. But this proves that everyone who talks about „X billion COBOL transactions“ is just blindly repeating internet rumors.
I later found the InformationWeek article Cobol Enters the 21st Century, that already mentioned 30 billion „COBOL“ transactions in 2003.
The COBOL report: 30 billion CICS transactions
So then I started digging for the source of „30 billion CICS transactions“ instead of „30 billion COBOL transactions“ and found that this number is often cited to come from „The COBOL Report“. This was hard to find, but after digging around on archive.org for a few hours, I nailed that reference down to this blog post by Scott Ankrum in January 2001 on cobolreport.com. That blog post cites „IBM, CICS Business Unit, Hursley. What’s new with CICS Transaction Server? – technically speaking. (training presentation). [2000, November]“ and I was absolutely not able to locate any resource for that. The section „CICS related Presentations“ on Bob Yelavichs personal homepage might have been close, but that was not fully saved on archive.org.
However, several sources (including Ankrum) say that the number of transactions had increased from 20 billion to 30 billion between 1998 and 2000 (the years vary a bit between different articles – 1998 and 2000 must be the correct years, given Ankrums and other references).
This increase was attributed to the rise of the internet. At the time, there were only about 30 billion web-requests per day, so what they were saying is that CICS handled a third of the internet-traffic of the entire world and the internet had zero growth outside of CICS related requests? I am inclined to doubt that.
Gartner: 20 billion
For 20 billion CICS transactions, I have found an excerpt from Gartner Group Symposium ITxpo: The Future of IT 1998 which says „IBM estimates that CICS systems run, on average, 20 billion transactions per day“. If Gartner had surveyed that number in 1997, they would not have phrased this like that in 1998. So we can conclude that this number doesn’t come from Gartner.
From the excerpt it is not clear who said this at the Gartner Group Symposium ITxpo. For all we know, this could have been said by IBM themselves. I am still trying to get my hands on a copy of the proceedings book. This one seems to be the only copy in existence.
Infoworld & CNN: 20 billion
The absolutely oldest, public source of the number I could find is the article „IBM, the world’s oldest start-up“ from the infoworld magazine, volume 20, issue 47, october 26th, 1998, which was even reposted on cnn.com.
This was only 10 days after the the Gartner Group Symposium ITxpo: The Future of IT, so I am very sure that the passing mention at this Symposium is where it all took off.
Walsh & Lamb: teleconference 1998
The cobolreport.com blog post by Scott Ankrum says
„One technique IBM uses to keep its customers‘ technical staff up to date is with a telephone conference call, using handouts distributed ahead of time. In one such teleconference (Walsh & Lamb, 1998), the speaker drew a comparison between a CICS transaction and an Internet web page „hit“.
This telephone conference was held on september 9th, 1998 – one day after the release of CICS transaction server 3.
The handout was nearly impossible to acquire. Even after I had found its original URL – it wasn’t even on archive.org. I had already given up on it. But investigating another hypothesis, I came across the usenet group bit.listserv.cics-l, where Christopher Frank had asked for that very handout on september 12th, 1998. In a later post on that usenet group, K K Leung cited Mark Granger, who had said „Sorry, I only sent this direct to Christopher as I thought he was the only interested party. Attached is a Word95 format doc.“ on October 13, 1998. So the file had been shared in that group, but Grangers message itself was not to be found on googles usenet mirror.
However I noticed the recipient of that message: CIC…@UGA.CC.UGA.EDU, which sounded like a mailinglist regarding CICS. So I went to uga.edu (The university of georgia), created an account on their listserv (thankfully, I was allowed to do so without an @uga.edu mail address), got access to the archive and voilá:
I got the cicswhitepaper.doc but it was all jumbled data. But the „BEGIN 644“ at the beginning was a very good clue: the file was only UUencoded, so I ran it through a UUdecoder and bingo: Dino DNA!
So after MONTHS of research on this, I proudly present to you the handout from the teleconference on september 8th, 1998!
„30 billion COBOL transactions“ is one of the biggest and most repeated marketing arguments for IBM, MicroFocus and others. It is attributed to Gartner, which makes it almost impeccable, because it is believed to be a result from a (Gartner) survey.
But we have seen, that the number is much less reliable: Due to a misunderstanding in a telephone conference in 1998, someone believed that it was an estimate by IBM. That supposed estimation was then mentioned at an event hosted by Gartner. And that passing mention is taken as justification to claim that the number came from Gartner ever since.
The number has been repeated and misquoted and adjusted for inflation so often that nobody seems to remember (or even care!) whether it has ever been true in the first place. Today it is reported as if it were fact.
But this leaves the question how many CICS transactions there really are per day. This has to be discussed in yet another blog article.