Tell the Difference Between These Three Pictures 1872

A reaction which displays a lack of familiarity with normal survey processing back in the days of punched cards.

Multipunching was the normal way of dealing with questions with multiple answers, like "Which of the following varieties of buscuits have you purchased in the last month?" If there was a list of 10 (or 12) items, these would be coded in one column with each row repesenting a yes or no for one item. The alternative, of coding a yes-no for each variety in a different column would chew up card real estate very fast. Where I worked back then (in a large market research agency) the production people greatly preferred a single card with multi-punching to a multiple card questionnaire. The "canned tools" of the day would have been perfectly happy with multi-punched data, and typically read cards in binary (bypbutting the hardware translation of hole-patterns to characters) to accomodate the coding.

I can imagine the background to this story. A guy with a background in survey research undertakes a project at a new employer or client business without considering that they are unfamiliar with standard practice. No need to consult with DP because he's not doing anything unusual. He sets up the coding in a particular way in order to facilitate the subsequent analysis. He's astounded that special-purpose code has to be written because standard analysis tools aren't available. If there were, the data entry and computer processing would have been cheaper (and less likely to introduce errors) than hand analysis.

--brian