Operational System of the ODE







	Ruhama Yitzhaki
	Surveys Division

	Tel. 972-2-655 3321
	FAX 972-2-655 3333
	e-mail: msruhama@pluto.mscc.huji.ac.il

General

1. Description of the system

The ODE system has handled the data entry of the census questionnaires. This
process included: saving an image of the questionnaire in the computer, automatic
conversion of the data into ASCII values in the database, human verification of that
data conversion, and coding of the alphabetic data.

The process is, in part, a serial procedure, in which the beginning of one stage
depends on the conclusion of the previous stage, but it was also partially
simultaneous.

The system operated at a site that had been specially set up for entering the
questionnaires and included a storage room for the paper questionnaires, a
scanning site, and four identical production lines, one of which also handled the
questionnaires in Arabic.

The scanning site was divided into four scanning units, and each scanning unit had
four scanners operated by two scanning operators. Each scanning unit was
directly linked to its own production line.

Within the production line itself there were stations for 17 employees: four keying
operators, four editors, two senior editors, a manager and assistant manager, and a
computer operator. The production line manager and the operator were also
responsible for the scanning unit linked to that particular production line.
Production line number 4 was set up somewhat differently (with additional personnel
for keying and geographic coding). In practice, the number of people employed at
the production lines was slightly less and their set up was somewhat different than
originally planned, due to the experience that had been gained during the course of
operating the system (explained in Section 5.2).

The smallest data entry unit was the Enumeration Area (EA). The working unit for
scanning, and consequently, for the production lines, was a group of EAs - all the
EAs in a sub-region. In total, the census took in 134 sub-regions, with each sub-
region having an average of about 60 EAs. Since scanning a sub-region took
around 3 days, the movement of the questionnaires of a particular sub-region, from
the storage room to the scanning stations and back into storage should have been
completed every three days for each of the four scanning units.

Following successful completion of the scanning, and successful performance of
tasks by the scanning server (scanning, FDO, OCR, compression and entering the
identification results into the database), the EA was ready to be transferred for
continuation of work within the production line. In order not to overload the
computer, the transfer of all the EAs for which work had been completed was done
at night.

2. Types of forms

The ODE system handled some two million questionnaires, which were composed of
five million pages, containing the records of six million persons.

The two million questionnaires consisted of several types of questionnaires:
questionnaires for households and questionnaires for residents of institutions.
These were divided into short forms, which were filled out by 80% of the
respondents, and long forms, which were filled out by 20% of the respondents.
The questionnaires for households were printed in two languages: Hebrew and
Arabic. Each household completed one questionnaire, except for households with
many persons, which filled out another questionnaire that was identical to the
previous questionnaire, and which was identified as being an "additional"
questionnaire.

The differences in the various forms does not affect the work of the scanners,
nor the work of the keying operators; they affect only the editing work.

Special preparations were made to enter the data from the Arabic questionnaires,
which covered two aspects:

The process of working with the Arabic questionnaires is different from
the process of working with the Hebrew questionnaires: since the
programs for identifying numbers are not suitable for Arabic numbers,
OCR was not performed on the Arabic questionnaires (with the exception of
the data on the first page of the questionnaire which was filled in with
identifiable numbers). All the data that was entered on the Arabic
questionnaires was keyed as text, i.e., "keying of fields", which is labor
intensive. Therefore, there were three additional keying stations in the
production line to work on keying the questionnaires in Arabic.

It was harder to hire suitable staff for entering questionnaires in Arabic
because workers were needed who were completely fluent in both
languages.

3. Physical conditions


	Data entry on the questionnaires was performed at a site which was suited to its special requirements, with an attempt to provide the best possible conditions for efficient work. The site included a storage room for all the paper questionnaires, which was close to the scanning area, four production lines, a training room, an experts' room, managers' rooms and a rest area.
	In placing the work stations on the production line, there was also an attempt to reach maximum efficiency; the production line manager, the operator and the senior editor all sat on a raised platform which enabled them to view and control the work process.
	Those with similar jobs sat in groups.
	In order to make the work environment comfortable, large tables were specially designed for the site which would allow the keying operator sufficient room for the keyboard and the computer mouse.
	The size of the computer screen was adapted to the different jobs (smaller screens for data entry, larger screens for the other functions) .
	The production line was equipped with "diffused" lighting which prevented reflections and glare from the computer screens.

4. Staff


	A total of 150 employees worked at the site. The work was performed in two shifts, each one 7.5 hours. Due to the nature of the work, which involved long periods of staring at the screens and very monotonous activity, there were four regular breaks during each shift. In addition, there was time for organizing both before and after the shift, so that the amount of time actually spent working at the site was about 6 hours per shift.

	The workers required to operate the site included: maintenance personnel, storage workers, scanners, production line workers, managers and experts. Each production line had a manager and an assistant manager, who managed the work on the production line. The site was permanently staffed by a software person, a geographical expert, and an expert in economic coding. Other experts came to the site from time to time. The entire site was managed by the manager, his assistant and the computerization manager.
	Management and computer operations were performed by Bureau workers. All other professions were supplied by an employment agency which was responsible for them in terms of personnel management.

	4.1 Staff recruitment
	Formal requirements were defined for each job. Those who met the requirements underwent two examinations, which primarily tested for accuracy and coordination. One of the tests took place at the computer and was specially designed for this system. In addition, each candidate underwent a personal interview. 58% of the candidates who took the test were accepted.

	4.2 Training
	Employee training was task-specific, based on the particular job one was assigned to do, with the exception of one day of joint training held for all employees to acquaint them with the census and the ODE system in general.
	Training lasted from three days for the more simple tasks, and up to one month for the more complicated jobs. Most of if took the form of on-the- job training.
	The production line managers served as the training staff for the various jobs. Each manager was in charge of training a particular task - keying, scanning and editing. Training for coding was carried out by the coding experts, who in turn received their training from CBS employees who are experts in this field.
	Once every two weeks, refresher sessions were held for the various professions, which were aimed at improving existing work procedures or providing a platform for questions and possible solutions, and to insure that any changes made were passed along to all employees in that particular profession.
	These meetings also occasionally included exams, in order to prevent incorrect procedures from taking root, to create a higher level of concentration at the training courses, and to insure that instructions were followed.

	4.3 Supervision and control
	Most of the employees at the site were charged with performing boring, tedious and repetitive work. This fact, combined with the relatively low level of education required, increases the importance of constant employee supervision.
	Three types of supervision and control were used in combination: computerized control, personal control by the production line managers, and external control.
	Computerized control: This primarily involved quantitative control and computerized monitoring of the rate and the output of each employee. There was a certain amount of quality control on the keying, and we planned computerized protection against common errors (such as entry of a mistaken EA number by the scanner, skipping keying items, etc.).
	Personal control: Since the number of employees for each production line was small, direct supervision of the workers was possible by the production line manager and his assistant.
	An additional control factor on the work of the editors was teaming a senior editor with an editor from another production line. The senior editor observed the work of the editor while he worked on a complete EA. For each editor, a feedback form was filled out which evaluated the handling of each separate editing problem by the editor being observed. Lessons were learned, raised and discussed according to the frequency of the problem - at the personal level, the production line level, or at "refresher" sessions.
	External control: While work was proceeding at the site, quality control was being carried out off-site. The results of this control was brought to the attention of the workers, by praising high-quality work, and pointing out the lessons to be learned from lower-quality work.

	4.4 Worker turnover
	One difficulty that must be taken into account is that the system is built for the short term. In this type of work environment, a high turnover rate is to be expected. In order to reduce the rate of employee turnover, wage incentives were offered to employees who remained until the job was completed. In total, the turnover rate was 20%, a relatively low rate. There were four recruitment and on-the-job training periods.

	4.5 Stand-by employees
	In order to prevent delays in the process due to employee absences, several workers were hired in the fields of editing and senior editing. One of them was always present at the site each day. The initial agreement with Manpower employment agency was that the stand-by workers could remain at home, but would come to the site as necessary, within two hours of their being called.
	As work progressed, it was decided that it was better to have one editor and one senior editor from among the stand-by employees always present at the site. The reason for this was that it became clear, based on experience, that during most of the time there was a lack in these professions due to employee absences, and, in order not to waste precious time, it was best that they be at the site from the beginning of the work shift. In the event these employees were not needed, they were sent home but paid for four hours of work. In addition to these two workers, there were another eight stand-by workers.

	4.6 Back-up workers
	After a month of work the need arose for several workers who were trained to perform various functions within the production line, thereby serving as back-up workers in the event that a particular profession became overloaded with work. Such an overload was caused due to simultaneous work processes, in all the professions, when an imbalance was created in several of the items arriving to be handled by the different professions.
	Back-up workers were trained as: keying operators, editors and economic coders, and there were eight of them.

5. Performance

5.1 Capacities

There was a gap between the scanning capacities and the work capacities
along the production line. In other words, there was a gap between the
capacities of the machine - which became steadily weaker, and the human
capacities - which steadily improved. In spite of the solutions that were
provided during the course of the work - balancing between production
lines by referring scanners to those production lines which were lacking,
and balancing the system by allocating days for scanning activities only -
these were only partial solutions, and the gap between the capacities led
to forced inactivity on the production lines.

In any system that is built according to a conveyer belt method, where
each phase depends on the previous phase, the capacity of the entire
system depends on the capacity of the slowest phase. The slowest phase
in this system was the scanning. The rate at which the scanning was
performed determined the work rate at the site, and therefore, also the
output, which was measured by the number of pages sent to the MF.

The scanning capacities (and the resulting output) over the five months
was about 44,000 pages per day. However, there was a continuous
improvement in the capacities throughout the course of the entire five
months' work, so that during the first month, the capacity was 36,000
pages per day, while in June, the capacity rose to 53,000 pages per
day.

5.2 Improving work efficiency

A shortening of the work time spent on each item was felt in each
profession, but differentially: In keying, improvement was continuous until
there was a reduction of about 25% in the work time; in editing,
improvement stabilized after about two months and the improvement was
about 20%; in economic coding, the rate stabilized after about one month
and the time spent working on each item was shortened by around 15%.
The differential improvement in the various professions creates bottle
necks and requires reorganization every so often. For example, as
mentioned in Section 1, despite the fact that each regular production line
was planned for four keying operators per shift, only three keying
operators actually worked per shift, and even then, there were periods
of inactivity.

5.3 Actual work time vs. with potential work time

There is a significant time difference between the potential work time on
an item and the actual work time spent on that item (time spent by worker
at the site, without any breaks).

"Wasted" time can be attributed to four factors:


1.	malfunctions
2.	a lack of balance between the flow of various items during the simultaneous work phase; for example, an EA from a religious area, where the number of children per household is large, will mean more work for the editor, in contrast with less work needed for economic coding. The economic coder will be forced to wait without any work to do until the editor has completed his task.
3.	lack of optimal staff planning; as mentioned above, differential improvement in the efficiency of those working in the various professions requires redeployment which would enable the system to respond to the new and changing requirements.
4.	the dependence on scanning, which was a phase that was almost impossible to improve.

The efficiency of the system could have been improved - that is, some of
the "wasted" time could have been prevented by separating the scanning
stage from the subsequent stages through the creation of a buffer
period prior to the start of any work on the production lines. Creating a
"stock" of work before the beginning of the work period would have
reduced the degree to which work on the production line depended on the
scanning, thereby insuring a reduction in underutilized work time.
Similarly, creation of such a buffer would have enabled gradual
absorption of staff and could have reduced the need for periodic
changes in staff planning. It should be noted that part of the problem
concerning underutilized work time stems from the simultaneous work, and
these will not be solved by the proposal offered above.