Page 198 - Orlicky's Material Requirements Planning

CHAPTER 9 System Records and Files 177

purpose cost something in terms of extra processing time, but for computer applications
in areas of high input-data error rates, by all means they should be programmed. A great
variety of this type of check is possible, and when carried out in a computer, it is the
swiftest and most efficient way of catching errors.
Internal detection during the actual processing of errors that got past the barrier dis-
cussed earlier is also an important system capability that usually can be programmed. It
is distinguished from diagnostic tests mainly by the fact that the checks are made against
the file being updated. An example might be a stock-disbursement transaction that pass-
es both input audit and diagnostic tests but is substantively incorrect, that is, reports a
withdrawal quantity that exceeds the quantity previously on hand. An entirely different
kind of test, called a test of reasonableness, sometimes also can be employed. For example,
if the use of a given inventory item averages 100 per period, a gross requirement of 1,000
or 5,000 for period X is almost certainly invalid. A human can spot and question such
absurdities immediately. A computer can be programmed to do the same. As far as a test
of reasonableness is concerned, the computer program, by applying this test, always can
flag the results that are suspect. The computer can tell the recipient of its output not to
use the information without verifying it first.
Washing out residues of errors that escaped detection by other means is a must if
the MRP system is to be kept from gradually (perhaps very gradually) deteriorating. It
always should be assumed that at least some small proportion of errors in input data will
penetrate the system despite all barriers and checks. These errors may be forever unde-
tectable as such, but procedures should be devised to detect their effect on system files,
and it is this effect that must be removed.
This is accomplished by means of various reconciliation, purging, and closeout pro-
cedures, which are analogous to writing off, periodically, miscellaneous small unpaid
balances in an accounts-receivable file. Examples of this type of procedure are the recon-
ciliation of planned versus actual requirements for an item at the MPS level and the clos-
ing out of ancient shop or purchase orders that still show some small quantity due.
It is a safe assumption that most MRP systems installed in industry contain some
(small, we hope) percentage of error at all times. This can be tolerated as long as such sys-
tem-resident errors do not accumulate. Even an accumulation at a minute rate eventual-
ly will smother an MRP system. As far as the washing out of error residues is concerned,
it is less important how soon after the fact the effect of an error is removed but all-impor-
tant that it be removed at some scheduled interval. If a certain level of residual error is
inherent to the system, it should be kept constant.
Equally important as the technical aspects of preserving data integrity—or perhaps
more so—is the training, discipline, and attitude of people. Those who contribute input
data to the system and those who use the system in the performance of their job must be
educated to the fact that a computer’s outputs cannot be better than its inputs. If an MRP
system is to be successful, management has to accept responsibility for convincing every-
one who interacts with the system that he or she has an important new role in “feeding”
the computer and thus helping to keep the system effective.