- Ed Igler a Man behind the Barcodes at IBM
- How the first magnetically stripped plastic cards utilizing the Bar Code were developed at IBM IRD in Dayton, N.J.
- Lemelson - Con Artist or Inventor?
- Managing with Store Level Data
- Early Grocery Store Level Efforts by IBM
- What we learned about people
- Where Did the Mirrored Bars for the Barcode Symbol on Aluminum Cans Come From?
- Recollections of The Early Days of UPC
What we learned about people
Or rather how scanners could make huge profits for stores that converted from mechanical checkout to scanners. Well not really….
When developing the specifications for the barcode and symbol one had to determine what would be an acceptable undetected error rate (UDE). In order to do this a simple survey was made to determine the error rate made by checkers keying in the price of each item. Men were sent to the parking lots of several stores chosen at random. These men would offer a shopper leaving the market with a large basket of items, a small payment ($10 as I remember it) to allow them to check their items against the receipt. What they found was that there was usually one or more errors in every one hundred items sold. Not surprising was that most of the errors were in the customer’s favor; I guess those in the store’s favor were caught by the customer and corrected.
It was reasoned that if a scanning system had less than 1 UDE in 1,000 items that the result would result in a modest gain for the market. The first time I was briefed on the specification orally I was told that to be safe, the UDE specification was raised to 1 UDE in 10,000 scans but when I received the written criterion it had again been changed upward to 1 UDE in 20,000 scans.
Before making our proposal to the Symbol Selection Committee, Dave Savir made an in depth mathematical model of my barcode & symbol. One of the conclusions in his published paper was one would expect a UDE of less than 1 in 100,000. WOW! I was overjoyed.
Our first scanner and software was very primitive by today’s standards. You must think back to 1973 when 128K of memory and a processor rate of a few megacycles was considered state of the art. Be that as it may, the scanner passed product test with flying colors and the UDE was expected to exceed our optimistic predictions and be actually less than 1 in 250,000. That is a long way from the 1 in 1,000 that the Symbol Selection Committee thought would be sufficient.
Needless to say, that is not the way it turned out. To understand the problem I must explain a little about the overall system of labeling, scanning, and checking. The bulk of the items would have the symbol printed on the label by a high speed press. When the item is scanned, the result is checked for parity of the characters and for the correct modulo check character. If these tests were successful the item number would be sent to the computer where a table lookup was made to retrieve the description and price. This was also a powerful checking tool because there were only a few thousand items on the file and the chance of a scanning error resulting in a match for the wrong item was slim.
Meat, cheese, and other random weight items were marked with labels printed in the store. These random weight items have the package price encoded in the symbol and a code number to get a description for the customer. The computer simply used the price in the symbol to charge for the item. There were no reasonableness checks, no table lookups, nothing. They were simply charged whatever the scanner read. To further complicate the problem, the symbol was often degraded by frost or juices.
Our first scanning systems hadn’t been installed very long when my manager, his manager, and two other men descended on Dave and me with a vengeance. Someone had been charged $99.99 for a pound of chicken. The “Golden Chicken” was born!
We tried to explain that overall the system was still operating with an UDE rate much better than the specified 1 in 20,000. That made no impression on them and I believe before that meeting ended or certainly not long after, we learned about “Platinum Pork” and some high priced cheese. Dave and I made a presentation to the Symbol Technical Committee 3 (STAC-3) to use one of the description digits as another modulo check digit on just the price. Dave had devised a complex algorithm for the added check digit that in conjunction with the original check digit made the validation virtually infallible. This was accepted. IBM quickly implemented the additional check digit along with some reasonableness checks on the price. Thank goodness…these changed allowed Dave and me to stay employed for a few more years.
The lesson we learned about people was that the cute little checkout clerk would be forgiven for charging $.98 for a $.87 item even though it would most likely go undetected. However, even in those days computers were considered infallible by the general public and a gross error which would never go undetected by the shopper would not be tolerated.