Ex Libris recommends loading a sample file (a "subset", in Ex Libris terminology) of test records into the software for initial evaluation of data loading, mapping, and indexing. Although FCLA can output a sample file of randomly selected records without library input, evaluating your data load can be much more effective if you select at least a portion of the records to be included in the test file. The recommended number of records for the initial subset is up to 75,000 bib records, along with all of their attached copy holdings, items, orders, circulation transactions, and patrons attached to the circulation transactions. Any records you flag can be supplemented with additonal records randomly selected from your database by FCLA.

Test record selection criteria:

Florida Gulf Coast University has developed a comprehensive checklist of records to select for the "subset" test load.

The following are additional suggested selection criteria for sample records:

Bibliographic records:
1. Long and complex bibliographic records.
2. Records with local headings.
3. Records with uppercase STAT codes. (Please note that STAT=D records will not migrate to Aleph.)
4. Examples of all NOTIS material types.
5. Samples of NOTIS Provisional records.
6. Samples of bib records with attached open orders.
7. Samples of bib records from each Processing Unit.

Copies:

8. Long and complex holdings records (be sure to include a number of records with MHLDs).
9. Samples of all classification systems used by your library (Class Type codes).
10. Records with uppercase STAT codes. (Please note that STAT=D records will not migrate to Aleph.)
11. If your library elects to migrate Z’d copies, be sure to include some examples in your Subset data.
12. Samples of copies from each of your NOTIS location, sublocations.
13. Samples of copy statements with copy level coded NOTE fields and User Messages (um=).
14. Samples of free text COPY NOTE fields.

Options for “Marking” Records for Subset Extraction:

FCLA offers two options for marking records for loading into the first Aleph test load:

  Provide FCLA with a consolidated list of NOTIS keys to extract: send FCLA a file of as many NOTIS keys as you want for extraction. Most libraries do not send list of 75,000 keys, so send as many NOTIS keys as you wish and FCLA will randomly select additional NOTIS records to complete a total of 75,000 sample records.

  Create Ticklers in test records: To facilitate sample record selection, FCLA has designated the use of two MARC tags: an 035 subfield a to contain a "tickler" code to indicate a sample record, and a 599 with a second indicator of "9" to contain any special notes about the test record. The tickler code should consist of your [NUC] + smp (for sample record) + one of the following codes:

b - for bib record testing
h - for holdings record testing
i - for item record testing
o - for order record testing
a - for authority record testing
all - for all records linked to this bibliographic record The last portion of the tickler can be a MARC tag number, so that sample/test records can be grouped together and searched in an index. The structure of the remaining portion of the 035 is up to each individual library. As long as the initial portion of the 035 is consistent, FCLA will be able to output the test recrods.

Examples: 035 |a(SNX)smpb 730

599 9|aselected by MWP to test indexing of 730

As you work with your database and identify sample records, add an 035 tickler field and a 599 second indicator 9 field.

Creating test records:
You may also want to create one or more "test" records for use in testing the loading, indexing, and display of records in the new system. These records might contain all possible fields and codes found in your NOTIS records, especially diacritics and any other special characters.

Subset Selection Recommendations from Phase 1 Libraries:

From UNF:

"The 2 things I'd add from the cataloging standpoint is to be sure to cover all classification types you want to test- all call number types including LC, documents, and any special, home-grown varieties. And second, make sure you have enough heading/authority control pairs to test linking (or not), e.g., traced and untraced series, earlier and later names, uniform titles, multi-level corporate names, all manner of subject headings, etc., in all systems, LC, MESH or any local."

From UF:

"When flagging records to put into your Test database, think broadly about how you'll be using the database."

Ask your entire staff to submit NOTIS numbers of their favorite records, especially the odd ones.

** get some very long records, either with long fields or lots of fields

** grab records that you've had to modify from OCLC

** grab local/original records

** grab records for items that are heavily used and likely to be checked out

Grab the ugly records:

** records with lots of items (New York Times, Newsweek, Books in Print)

** records with multiple copies and/or holdings statements

** make sure you've included all the MARC tags among your selections

Think ahead to using your Test database as a staff training tool, and grab records that will make good training examples.

** grab a numbered series with a serial record and lots of monographic records (20-30)

** grab all your Shakespeares

** grab all or lots of records from some lesser authors, where you'll have somewhere between 51 and 500 records (Dickens, Asimov, etc.)

** grab titles that have records for multiple print editions, separate print + online records

** grab lots of records (between 100-500) that can be retrieved from a keyword search to play with sorting features

** grab all title change records from some of those annoying journals"

** All the Shakespeares can be overwhelming, or very slow to load some days