What are the difficult issues while implementing an automated data entry software system?
In the first few weeks the operators who were only using the keyboard before may find it strange and difficult to accept that they have to work a bit differently from now on.
They need to put the paper stacks into the scanner continuously (usually 50 sheets fit into the feeder at once) and they need to take the papers from the output tray. They have to make sure on the recognizing computer that it processes the proper stacks. Actually they need to spend most of their time in front of the corrector surface and compare the scanned field images with the characters recognized by the computer and correct them, if necessary.
In general, the users overcome the initial difficulties in 4 to 6 weeks, in case of larger systems, and then they would object if they had to work purely manually. By this time they understand and experience that the computerized automatic data entering facilitates their work greatly.
In this work method we do not really mean "data recorders" but a staff which confirms the data entered by the computer instead of making data entering. And it means less work load for the staff.
In which situations is it purposeful to implement a data entry system?
In cases when the computerized system is much quicker and works with less error than the data recording personnel.
Another important factor is the "critical mass", which means that the task should demand the processing of relatively many forms. Ideally many forms of the same type must be processed.
Also, it depends on the form’s data content whether there is a point in using the computer. For example, a form that only contains check boxes and bar codes is the best for the computer, it can read the form with 100% accuracy and the data of the recognition do not need to be validated either, it means quick processing. The other extremity is a form containing manually written words in block capitals and small letters with accents from various people (e.g. when names and private addresses are collected for the buyers’ club of a supermarket), the rate of recognition is rather low here, only about 60-70%, the correction of the errors may be so time-consuming that it is more purposeful to enter the data manually. In the actually suitable real applications check boxes, bar codes, printed letters and numbers (readable in 100%) and manually written numbers (recognizable with 98% accuracy) are found.
A frequent solution is to record the parts which the computer can only recognize with difficulty (such as parts in current handwriting) by the traditional manual method as entering is quicker even so because the computer processes the other parts.