Auto Sort Exercise
April 1963: Martins Bank embarks on an experiment that is to have a profound effect on the way in which payments from bank accounts throughout the United Kingdom are to be computerised. This is another first for the bank, as none of its competitors has yet so regularly processed batches of cheques using the new reader-sorter technology.
Following a great deal of research, much of it by Martins’ own Ron Hindle, the Committee of London Clearing Bankers publishes “Requirements for Automatic Cheque Processing” in 1962, but Martins is first to seize the initiative and convert words into actions.
“The Pegasus Computer System installed in Liverpool in 1961 was in use for branch book-keeping before any other bank announced similar developments. The I B M Reader/Sorter was installed in London in the same year and work is in progress to marry the current account and the clearing operations in order to provide an integrated accounting system”
Sir John Nicholson, Bart CLE JP - Chairman of Martins Bank - Summer 1963
To keep the staff of the Bank informed of these important developments, Martins Bank Magazine publishes the following article in its Autumn 1963 issue. It gives a glimpse into the future of the ordinary bank branch – and crucially, the machinery and processes that everyone will have to adopt to cope with the exponential growth of payments by cheque that will characterise banking in the 1970s.
AUTO SORT EXERCISE
early in April branches were told of a further development in the Bank's electronic programme. The automatic cheque sorter installed in the Clearing Department at 68 Lombard Street towards the end of 1961 was ready to embark upon an exercise, the first to be attempted by a bank in this country, involving the sorting of part of the Bank's general clearing. Branches have seen the end product of this exercise in the form of a batch of cheques, received separately from the bulk of a day's clearing, with the amount of each cheque printed beneath the signature in magnetisable characters, and they have played their part in the exercise by confirming the correctness of the encoding.
To find out what happens to the cheques before they reach the branches we visited the Clearing Department at Lombard Street and were guided through the various stages by Mr W. S. M. Wilson, Assistant Clearing Manager. Mr Wilson began by explaining why only part of the clearing passed through the sorter. First of all, less than 60% of our cheques at present bear the encoded branch sorting code number so that by passing all the cheques through the sorter nearly half would be rejected and have to be sorted by hand. Secondly, an experiment is always conducted on a comparatively small scale and additional encoding machines would be necessary to handle all the clearing.
We then followed the course of one of the batches of our cheques received that morning from the Clearing House. The usual process of listing the cheques by amount to agree the total given by the remitting bank was followed but using one of the Department's encoding machines. The amounts of the cheques were keyed and listed on the paper roll in the usual way but after each cheque had been listed it was placed in the compartment to the left of the keyboard where the amount, held in the machine's 'memory', was automatically printed in the E13B magnetisable characters under the signature.
The listing and encoding completed, the total amount of the cheques was compared with the remitting bank's figure and, as it happened, found to be 10d short. Once the error was located—a cheque for £12.15.10d had been listed as £12.15.0d—the correcting method was explained and demonstrated. An encoded amount can be read by the automatic sorter only if preceded and followed by the 'amount symbol1 (see page 30): remove one of these symbols and the sorter will ignore the amount. For the purpose of the exercise, the sorter's ignoring the amount of a wrongly encoded cheque is preferable to encoding it again and this end is achieved by deleting the last amount symbol. The magnetisable property in the encoding ink is ferrous oxide and a special magnetic lead pencil which can deposit a further coating of oxide is used to obliterate the last amount symbol. The cheque incorrectly listed as £12.15.0d was treated in this way.
THE SORTER'S CAPABILITIES
The batch of cheques was then ready for the IBM Reader/Sorter, to give the sorter its full name, but before the various processes began we learned something of the machine's working. The cheques are fed lengthways into the mechanism at the right and pass the 10-cell reading head which interprets each magnetic character as illustrated overleaf. We found it hard to appreciate that this reading head makes seven separate readings of each character as the cheque passes by at 15 m.p.h. It does, however, give an idea of the capabilities of electronics and the speed with which cheques can be handled. The sorter is capable of dealing with nearly 950 cheques a minute provided they are of uniform size and in new condition. Naturally, this rate is reduced when the cheques are of various sizes and in the crumpled state often met with in the clearing.
From the reading head the cheques pass between metal strips, one of which runs to each of the sorting pockets. The sorter selects the appropriate metal strip according to the characters read and the reading or sorting task it has been set. There are thirteen sorting pockets, ten of which are numbered 0 to 9, two for special sorting marked A and B, and the last, marked R, for rejects. With mind still boggling at the ingenuity of man in creating such a machine, we watched the first run or 'pass' as it is called. On this, the reading pass, the sorter added the amounts of the cheques as encoded and produced a total for balancing against the earlier listing. All the cheques passed to one pocket except one 'reject' which was deposited in pocket R: this was the cheque for £12.15.0d which had been wrongly encoded. It was necessary to add this amount to the sorter's total to balance the batch.
Before coming to the first sorting pass we had to learn more about branch sorting code numbers. Following our Bank's number (11) are four digits which are counted from the right. The first digit is nearly always 0 and is not used in electronic sorting. (Its value is appreciated by the human eye, however: Sea Houses branch— 11-59-71—is quite distinct from Seaham—11-5770). It is on digits 2, 3 and 4 from the right that the sorting rests and experience has shown that the speediest method is to start with the third digit. The first sorting pass was made, therefore, on digit 3 with the sorter instructed accordingly. Each cheque with number 1 as its third digit dropped into pocket 1, number 2 into pocket 2 and so on. The 'rejects' this time were those cheques on which no sorting code number appeared at the foot in E13B characters and which required manual sorting.
Before removing the partially sorted cheques from their pockets for the second sorting pass the machine was fed with sets of yellow and red cards, each the size of an eight inch cheque. There was a card of each colour for every branch and the purpose of the yellow card was explained to us first. Each carries a branch name and en- coded sorting code number; in addition it is encoded with an amount of nothing—a row of O's. The need for an 'amount' lies in the machine's built-in logic which would cause the rejection of any item not bearing an amount on the final reading pass. Fed through the sorter, the yellow card drops behind the cheques and from then on it follows the cheques drawn on its branch so that when the sorting is complete each branch's batch of cheques is separated from the next by a yellow card—so much more convenient than having to search for the last cheque!
The red or 'auto-total' cards, which are fed in behind the yellow cards, play a vital role in the exercise. Once the cheques are sorted into branches, amount totals are required for entering on the Auto-Sort exercise response slips forwarded to branches along with the cheques. Each red card, lying behind the yellow card at the back of each branch batch, is encoded with the amount of Id and on the final pass—a reading pass—the machine is instructed to throw out a total of the cheques it is electronically listing each time it meets a single amount of Id, and at the same time to discard the red cards. It is further instructed to ignore the amount of Id in its additions.
We watched these final stages—the feeding of the yellow cards, the red cards, the last two sorting passes on digits 4 and 2, the reading pass giving branch totals and the rejection of the red cards. All that remained was the manual attaching of the exercise response slips to the appropriate bundle of cheques and the removal of the yellow cards. The exercise has been a success and the eventual link-up with a computer can now be envisaged. This will drastically cut the operating time as a number of the passes, at present necessary to the exercise, will be omitted. The first reading pass will be superfluous—the amount will be fed to the computer as the first sorting pass is made—and by adapting the sorting pockets most cheques will be both read and sorted into branch batches in only two passes and the remainder in one further pass. In a few years' time we shall have ceased to marvel: electronics will be a part of the daily life of every one of us in the Bank.
The principle of electronic reading, showing how characters printed in magnetic ink are sensed and interpreted.
The multi-channel reading head breaks up the characters into 10 horizontal elements. Seven separate readings are made as the character passes under the reading head, so that each numeral or special character is divided into 70 blocks, or sensing areas. The presence or absence of magnetic ink in each of these sensing areas is recorded in an electronic register, which therefore builds up an image of the printed character. Powerful logical circuits compare this image with the perfect character, and with the thousands of permitted variations. Some variation is inevitable, and the machine's usefulness would be very limited if only perfect reproductions were accepted. It therefore rejects only those images in the electronic register which could be related to more than one of the 14 possible characters. The characters which make up the E13B typeface have been carefully designed after exhaustive analyses of comparative shapes. Each is as different as possible from all the others, yet retains its legibility for the human reader. This style has been adopted by all British and American banks for use in machine reading systems.