Rummaging through some old files, I ran into an artifact originating from that time before computers, although this was still in use long after computers became available. Such is the nature of progressive automation that we resist it for long after it first becomes feasible. Eventually the old ways are tossed aside, leaving piles of obsolete artifacts. For some reason I kept one.
The artifact is called a keysort card, also called the McBee Keysort. Its purpose was to record data in such a way that statistical reports could be produced that summarized data across hundreds or thousands of these cards. All of the work to do this was performed manually.
In the mid-1970’s I worked for Westchester County (NY) Department of Health. My role was to perform inspections, record observations and file reports. All of my observations were recorded on this card. Most important to the County was to justify how these activities could be reimbursed under various state and federal programs. To do that they had to extract pertinent records from massive piles of same and summarize according to the fields controlling the extraction. In filling out the card I had to check off each field that applied to the inspection. If I was responding to a request from a day camp to get a water sample for bacterial testing I would check ORIGIN/REQUEST, SAMPLE/WATER, and TESTS/BACT. The card was three parts such that the carbon backing on the first and second parts would copy the writing onto the next part. The third part was cardboard with perforations as shown below.
When the cards were submitted to the central office, they only needed the cardboard part. The statistical clerks would take each card and cut a V notch wherever a card had a check mark thus turning a circular hole into an open notch. Do you see where this is going?
Cards were collected by date into batches. For each batch, they might want to count the number of water samples collected for bacterial testing. Grabbing a batch of cards, the statistical clerk would insert metal rods like a knitting needle into the holes corresponding to SAMPLE/WATER and TESTS/BACT and then shake the batch until all of the cards with V notches in those positions would fall out. I am not kidding.
Now they had a pile of cards that they would manually count and record the result before moving on to the next batch. This corresponds to an SQL query something like:
SELECT COUNT(tests) FROM keysort WHERE sample = "water" AND tests = "bact" ;
This process required two full time statistical clerks who had this as their only task, given the number of cards and the number of reports that were needed. Eventually this process was replaced by a computer application. I was gone by then. You see I noticed that the computer programmers and analysts were all driving new cars while the health inspectors were driving old jalopies. I never looked back, until I found this old keysort card.
