At the University of North Texas we have been using VTLS for periodical
check in for about 2 years.  VTLS Serials Module uses the USMARC Format
for Holdings and Locations as its basis for check in.  We have come to
several conclusions:

1.  Collection of initial data regarding the publication patterns takes a
considerable amount of time.  The maintenance of this data is not nearly as
difficult.  Most change frequency at the beginning of a year or a volume.
We approximate 1% of the 3,000 titles we have automated have changed
frequency this year.

2.  The use of the USMARC Format (or predictive check in) in an automated
system requires a higher level of staff than was previously necessary for
check in on a kardex.  The staff person has to understand the USMARC Format,
the automated system, and the routine of check in.  It is also necessary to
assign the maintenance of the predictive patterns to someone to maintain
consistency.

3.  We would recommend against expecting vendors to provide the data for
predictive check in.  It is preferable to have the issue in hand to set up
the initial records.  It is absolutely necessary to have the previous pattern
of check in.  The vendor can not provide this information accurately because
they never see the issues.  Even relying on the publishers stated frequency
can cause problems.  The frequency "quarterly" can mean simply 4 issues per
volume, 4 issues published during designated months, or 4 issues published
with seasons.

4.  We have found that predictive check in is unsuccessful with any irregular
publication pattern.  There are also problems with any titles received
erratically.  You can't predict a slow boat from China!  The diversity of
your collection may be the biggest factor determining the success of predictive
check in.  Predictive check in works best for North American periodicals with
regular publication patterns.  Anything outside of this tends to be more
problematic, such as the sciences and some areas of the fine arts.

5.  The "value" of maintaining the data for predictive check in is not really
of question.  If the system is going to work the data must be current.
Although predictive check in would not be our first choice for technical
processing of serials, our patrons seem to be pleased with the public display.
The vast majority of periodical titles will need no more maintenance with
predictive check ithan they did on a kardex.

Hope this information helps answer some questions and concerns.  If you have
specific questions please feel free to contact either of us.

Lisa A. Macklin
Serials Records Librarian
LMACKLIN@UNTVAX.BITNET

Margaret E. Galloway
Automated Serials Librarian
IH58@UNTVAX.BITNET