An experimental analysis of critical factors in automatic data acquisition through bar coding

The technology of bar coding has been in existence for nearly forty (40) years but has only recently found much application in modern industry. This fact is attributable in part to the evolution of the bar coding symbology itself (of which there are at least 16 in use today), but to a larger extent to the technological advances that have greatly improved the ability to both print and read bar coded symbols. There remain, however, a number of critical factors that are thought to impact the success or failure of a bar code system.

Although several methods for measuring the success of bar code system are certainly plausible, the most appropriate or revealing index of success is identified as First Read Rate (FRR), for a low FRR will virtually guarantee user rejection of the system in favor of a more traditional yet undoubtedly slower and less accurate method of data collection or reporting. But while the importance of a high FRR is generally accepted, the factors or parameters that impact FRR are to a large extent still unknown.

Defined herein are two major categories of bar code system parameters. There are:

1. (1) GO/NOGO Parameters

2. (2) Level of Success parameters

Of the “Level of success” parameters, the quality of the printed media or print technique is often purported to be the single most important criterion in determining FRR. Other potentially significant contributors to the success of a bar code system, however, include the bar code application (defined herein by label length), and the human variability of the operator(s) that must use the system.

This text outline an experiment designed to characterize the impact of these three parameters on the FRR of a bar code system. Then a mixed-effects linear model is defined, and factorial analysis of variance (ANOVA) techniques are used to analyze the results of 4800 attempted “reads” which represent the FRR data for every possible combination of four (4) print techniques, three (3) application/label lengths, and two (2) randomly selected operations.

Contrary to the assertions of many “experts” the results of this experiment lead to the conclusion that the operator effect and interaction effects between operator and the other experimental variables are likely to have the greatest impact on a system's FRR. This conclusion suggests that success of the overall system is tantamount to success in controlling the operator variability, and that more attention should be given to the definition of human factors such as operator training than to the specification of system hardware — as is so often the case.