Tuesday, April 23, 2013

My Life With Technology - Chapter 1.2

by Bill Holmes

This next chapter in my continuing saga with technology from the stone ages has to do with the introduction of credit cards for the masses. Up until the early 1970's, universal credit cards were not a fact of life for most families. There were store credit cards, especially for high dollar stores, although "card" may not be the right term. They were more like credit accounts where the privileged few just said "put that on my account" or "put that on Daddy's account". I never got to say that. There were revolving accounts at some retailers, usually reserved for large purchases like a new refrigerator. I think Sears and Penney's were beginning to issue store cards. Weren't the old Sears and Penny's cards about a third or a half the size of a normal credit card?  There were gas cards that were limited to the issuing company's (Gulf, Shell, Texaco) gas stations and products. Diners Club and American Express introduced more general cards in the 1950's. Those were for the rich or business travelers and were not accepted outside the transportation, lodging and dining industries and not by all of them. That all changed in the late 1960's and early 70's. BankAmericard (now VISA), soon followed by Master Charge (now MasterCard), began introducing a generic credit card. One not attached to a specific store or industry. More importantly, one not confined to the rich and famous. They franchised banks around the country to issue and administer the cards. If you're old enough, you may remember that there were mass mailings of actual credit cards. There were also mass mailings of applications that the banks were not prepared to process so most were approved. What could possibly go wrong?

Somewhere around 1969 or 1970, First National Bank of Atlanta (FNB) became the lead BankAmericard bank for the southeast USA. That will figure in again when I get to chapter 3.0 of this saga. For now I'll relay the events of a major bank in a major market that was overwhelmed by this new consumer credit card.

I think FNB mailed a credit card to every account holder they had. If we had your name and address you got one. No need to have good credit or even a positive balance. No need to call to activate the card, just start using it. Guess what, some of those new cardholders didn't pay their bills. Other cards fell into the wrong hands. Either by people stealing mail or rummaging through trash. Often the people that didn't want them just threw the cards out. Blind mailings back then weren't quite as bad as they would be now because not every business accepted these newfangled credit cards. Of course the bad guys found every establishment that did. There were significant write-offs.

The franchised BankAmericard banks had to issue and process their own cards. There was no national card processing center. I'm sure Bank of America gave/sold us the software to do that, but it was less than stellar. There were technological limitations back then too.

Enough about the early credit card fiscal problems, let's move onto the the side I know about, the processing of those accounts and transactions. It was a mess. At FNB we were no where near ready to process this application. There were no new computers installed or on order and little or no training.  I can only guess that the powers that be thought credit cards would be a small, slow growing, application or that we had tons of excess computing capacity. Neither assumption was correct.

Even with  low transaction volumes, the BankAmericard application took time to process. Time we didn't really have to spare. There was also little or no infrastructure in place. In many ways we were going backwards. Typical, either the mucky mucks forgot to ask if it was possible or feasible or the lower level manages were afraid to say they couldn't absorb the work. They never asked me.

I'm going to describe what the computer and support departments had to do to process the BankAmericard application. Realize that there were many steps by several departments before and after our part. None of those departments were any better prepared than we were. I'm sure I'll leave something out or get something wrong but I think you'll get the gist.

If you are old enough, you remember getting a punch card in your credit card bill. That was the part you mailed in with your check to pay your bill. Do you remember that it said "do not fold, spindle or mutilate". There's a reason for that, that card/statement had to be processed by humans and computers to get your payment to the right place. You may also remember that the bottom copy of that credit card slip you signed at the store was card stock. That also was sent to the bank by the merchants to be processed. By "sent", I don't mean electronically transmitted over the internet. I mean physically taken to the bank or mailed, snail mail, not email.

51 Column Receipt
80 Column Receipt
Let's talk about the individual transactions first. You may find this hard to believe, but sometimes they got "folded, spindled & mutilated", they also got stapled, torn, stained, taped, wet, glued and any other thing you can do to a piece of card stock. They were often impossible to read. They came in two different sizes, the regular 80 column card size and the compact 51 column size. That is an enormous logistical problem. Almost every keypunch, card reader and tabulating machine is set for 80 column, usually pristine, cards. Now we have 80 and 51 column cards that have been abused. They had to be separated to be processed. There were attachments for the various machines to switch between the two sizes but it was far from automatic.
Credit Card Reader

Those original cards found there way to our Keypunch Department. Once there they had to have the account number, merchant number and transaction amount punched into the card. This was not only a big increase in workload, but a big change in input source documents. Now you are keypunching data into an original source document rather than punching a blank card. These documents were the third copy of a three part form in the days of carbon paper. Often this third copy was not the easiest to read. Besides faint or smudged credit card images, the dollar amounts were hand written. Guess what, some people have terrible handwriting. Others don't press very hard when they write. What happens if keypunch makes a mistake? Unlike blank stock cards, you couldn't just throw the botched receipt away. We had to find a way to cover the erroneous punches and fix it. Some company invented itty bitty silver mylar tape patches to cover the bad punches. We bought them by the case. These patches fixed the immediate problem but they caused many others. Tape comes loose, either completely or at a corner. The sticky side of the patch was over a hole which could stick to another card or pick up dirt, dust and other crud. All this could cause jams and misreads in the equipment that had to process these cards. An IBM 084 card sorter ran at 2,000 cards per minute (CPM), the IBM 2501 card reader or 2540 reader/punch on the computer read 600 CPM. As you can imagine, the tolerances were pretty tight. A piece of tape, no matter how small, in the machine guts will cause a problem. You can mangle quite a few cards when a jam occurs at those speeds.

So, occasionally the original document is completely botched by keypunch or mutilated somewhere else up the line. Too much for tiny patches to fix. We had a solution for that. We had a couple Xerox machines (all copiers were Xerox's regardless of the actual manufacturer) that spit out cards rather than sheets of paper. Put the bad card in the copier, a new card emerged, either 80 or 51 columns. The now pristine card went back to keypunch to have the relevant data entered. Of course the copy of a mangled card was often pretty hard to read too. For some reason that escapes me, that copier was in the production control department and not the keypunch department. Those departments were next to each other, but it still was farther away from the main user than needed. There was a procession of keypunchers through our department all day and night long.

When keypunch finished a batch of the charge slips they were sent to Tabulating (Tab) to be listed and added. If it didn't add up to the correct total someone in production control would check it. Then back to keypunch, back to Tab... Round and round until it balanced. Even if it was just one item that needed to be corrected, the whole batch had to be listed again. This was a bank and there had to be proof that everything balanced at every step. Now every time these cards were handled it was another opportunity for a jam or a correction patch to come loose. Sometimes we'd correct the one bad card in a batch only to have three or four errors on the next listing.

Eventually everything would balance and it would be time to actually run the BankAmericard applications on the IBM 360/40 computer. All these cards that had been handled multiple times now had to be read on the computer card reader and written to disk. One pass for the 80 column cards and one for the 51 column cards. Another opportunity for jams and misreads. We hardly ever missed an opportunity to have problems. The 51 column cards were especially troublesome.

After finally making it to the computer disks, the transactions were sorted for two part processing. There was merchant processing which credited all the transactions submitted by particular merchant to their account. That amount would eventually wind up as a credit to the merchant's checking account. The other process was for the individual card holder accounts. This was where your purchases were added to your account and payments were deducted. Any interest and fees were also calculated. All kinds of reports were produced for both the merchant and card holder accounts. The biggest pain was the customer statements. There were ten statement cycles a month, one every three days. Of course with weekends we actually had about 21 days to print the ten cycles. We absolutely loved the days we didn't have to print customer statements.

OK, we suffered through card based transactions. So what did we do for BankAmericard statements? Well of course we produced cards that you sent back with your payment. It took a multitude of people and machines to produce those statements. Here is the life cycle of a BankAmericard statement in 1970.

I mentioned last chapter that the IBM 2321 Data Cell, aka Noodle Snatcher, had been designed by Rube Goldberg. Well Rube or one of his close relatives designed the BankAmericard statements and their processing. This will get a little complicated and confusing, so if you get a headache just skip to the next paragraph. The statements were two up continuous form 80 column card stock. Each row of cards had a small strip separating it from the previous and following rows. The statements and separating strips were perforated on the top, bottom and sides. I have absolutely no idea why those little strips were there. The blank statements has a sequence number pre-punched and printed on them. When it came time to print the statements the operator had to try to line up the first row then enter the sequence number on the computer console. Initially there was no print test pattern to check the alignment. There was a single cycle button on the printer so you could print just one line at a time. If you got a misaligned statement way back then it was probably one of the first few printed. These statements were a pain to print. Besides getting them started, with all the perforations and the heavy stock they tended to jam. They were going at 2,000 lines per minute (LPM) so once again jams aren't fun. They didn't stack very well either. When we finally got those statements printed they had to be bursted and collated.
Forms Burster -
not the Rube Goldberg model
Rube Goldberg's family was again involved in the special burster/collator we had to get to separate the statements, remove the little strips between the rows and make a two up form into one stack of cards. Once all that was done, we sent the now separated statements back to the computer room. They were then fed into the punch side of the IBM 2540. The sequence number was read by the punch, compared to the statement file (on disk or tape) and the account number and minimum payment were punched into the statement. After about two days we found out that Rube Goldberg's burster/collator let some of those separator strips slip through. That caused problems. We eventually sent the statements to Tab to run through one of their machines to find all the chads. Of course the Tab guys loved having to clean up the garbage for the computers. The exact steps that Tab did to clean up the cards changed and evolved over time. Eventually those statements came back to the bank to be inputted as payments, but usually only after they had been "folded, spindled and mutilated" or worse.

IBM 2540 Reader/Punch  
It was and is complete stupidity to have an application that depends on slooooooooow card readers, punches and printers attached to the computer. We were running a very expensive computer at the 600 or 2,000 CPM or 2,000 LPM speeds of mechanical peripherals. That was optimal speed, not counting card and print jams. Collect the data on cheap devices and let high speed computers do their work at high speed. Now of course most raw data is captured at the source and transmitted digitally to the proper processing center.  

Once those source document charge receipts finally got entered into the computer for processing they had another journey. Are you old enough to remember that you used to get those credit card receipts back in your BankAmericard or Master Card statement? Kinda like you used to get your cancelled checks back too. Well, all those credit card transactions had to be sorted by account number. BankAmericard account numbers were thirteen digits back then they are now sixteen digits. You could eliminate the first digit, four indicated BankAmericard, still does. The next three digits were the issuing bank number, FNB was 327. That still left nine digits. Then you had to sort these folded, spindled, mutilated and patched documents into account number sequence so they could be sent back to the customer. That's another of the ways the old Tabulating Department stayed relevant. They used a combination of sorters and collators to do that. Each day's transactions were sorted then merged with the transactions from previous days. When a statement cycle dropped, those statements were merged with the sorted transactions in preparation for mailing to the customers. As you can see, those cards were run through various machines many times.

The whole BankAmericard application was a cluster f**k. Here we were with a whole infrastructure that was geared to processing checks. We were experts at MICR (magnetic ink character recognition) document processing. We processed over a million checks a day. Punch cards were for the relatively few documents that couldn't be read on the MICR equipment. So, what do we do, we introduce a punch card centric application. Just when punch cards are becoming obsolete. We weren't the only idiots, every credit card receipt back then was two plies of flimsy paper, two plies of carbon paper with a card stock bottom copy. I'm sure the Master Card banks were struggling too. It took as much or more time to process a few thousand credit card transactions as it took to process a million checks.

As complicated and convoluted as what I've described was, it was no where near the full story. There were many more steps before and after what we did in the data processing departments. There were whole new departments created. I have some general knowledge of what they did but not all the details.  The fact that anyone, merchants or cardholders, ever got a statement let alone an accurate one, is amazing.

I have no idea how long it was before FNB made any money on their BankAmericards. Besides all the startup costs for issuing and processing the cards, they had big write-offs because of the mass mailings of unsolicited cards to everybody with a pulse.

Over time the processing improved. Some of it was refinement or elimination of the steps. Some of it was getting more experience. Some of it was improvements in technology. Nowadays you can make a credit card purchase, receive a bill and make a payment without generating a single piece of paper (or card stock). We've come a very long way in 40 years. I enjoyed most of my time working back then but no one enjoyed those early days of BankAmericard processing. Let's call it a learning experience. Isn't there an old saying about that which doesn't kill you makes you stronger? Many of us got very, very strong in the first few months after consumer credit cards were introduced. Back in the day, technology wasn't for sissies.

If you found this account confusing, disjointed, complicated, puzzling, unclear and baffling then I made my point. Imagine how we felt trying to do it. I hope you also found it interesting.

I think the next chapter will be about my first experiences with check processing and the technology used circa 1970. It wasn't pretty, but it was many times better than the early days of credit card processing.


No comments:

Post a Comment