Remove restriction on number of Primo facets value from CDI
CDI facets are artificially restricted to a maximum of 20 facet values, including for content type, collection, language, and topic (subject). For the Journal Title facet, this is even lower at 10 facet values. The only time more facet values will be returned from CDI records is when local records share the same data and are returned in the results set.
There is clear evidence from our Primo Analytics data that that our users value this feature, with an average of 1,550 to 1,750 facet filtering actions per day over the last few years.
The result of this restriction is a poorer experience for our users with CDI. Our users will rarely see and be able to use facets which have fewer records in CDI only, such as Market Research content, subject terms associated with newer research, and poor visibility of diverse languages. Content from more niche smaller source collections are completely obscured by bigger aggregators in the source facet.
Also, with the greater number of results returned by design in CDI, the ability to use facets to filter results and reduce the noise is even more important than ever. But this restriction removes the ability of users to even seen this content is available in the facets and pick it out by including active filters.
Ex Libris states that this limit was imposed due to performance issues and that adjustment is an enhancement request, despite that this is a significant downgrade in functionality from the facets available in a Primo environment with Primo Central Index records (see case #00823948).
The facet number generated varied with environment but is 2,000 facet values for Multi-Tenant Back Office by ‘Number of Top Hits for Facet Creation’ in Search Engine Configuration. Ex Libris should not be overriding this for CDI and blended searches, as this is not a reasonable solution to alleviate poor performance on CDI.
Outcome: Abide by the environment number of top hits for facet creation, of at least 2,000 facet values, and remove the artificially low limit on CDI facet values.
> 500 Maximum Number of Displayed Facets
We're on Primo VE and many facets where alphabetical sort would be much better / easier to grasp than number of hits, but we can't switch to alphabetical display because it only can be used if we're sure the facet won't have more than 50 entries.
Stacey van Groll commented
To continue the discussion of facet values for local records, this is yet another parity issue between Primo using Back Office and Primo VE managed via Alma, and I did not realise until recently just how limited VE was.
Our Back Office is currently set to 500 Maximum Number of Displayed Facets. We have never had a complaint on either this limit, or desire wanting even more from users, and I have never felt hindered by this high level of data as a systems librarian. I have also never noticed any performance delay with loading these values, and they appear pleasingly immediately at the time of results returned or by selection of 'More'.
Without being allowed to see more facets in UI, there are issues such as when there are few resource types in the set. For example, if we have a facet for Library Guides this is likely to only be one records in millions. In BO, we can easily expand the facet list for the Resource Type set to size, and find the Library Guide (1). And expanding the Subject Term or Author or Language facet etc, there are the expected 500 entries shown.
The ability to see this level of data as a feature of UI flows beyond just selecting to include or exclude a single facet in a one-time search. Users can use the facet selection to tag resources easily in bulk into My Favourites, can use the locked facet feature to do subsequent more limited searches on the set without having to perform another Advanced Search with complex search lines on specific fields to attempt to target the data (and this expects that they know what to search and are not helped by easily viewing the entries in the facet), and they can set up a Saved Search and Alert specifically by this facet to quickly re-run the query including or excluding facets, and get regularly email updates with these facet value selections.
Why is there this limitation in VE? Is there some fundamental design flaw causing such performance issues that Ex Libris decides that this part of Primo UI must be is drastically and artificially limited to overcome it? If the argument that this is acceptable is the common one of relevance ranking, I would say that not all problems are solved by relevance ranking, such as the issue of long tail results from CDI meaning that Sorting results is not possible, and it does not remove the value of other features such as facets to help with this.
"Number of Top Hits for Facet Creation – This field specifies the number of top-ranked records the system will use to create the list of facet values that are displayed in the Refine My Results section of the Front End. Currently, this field is no longer configurable and is set to 2000 internally. For CDI, all records in the results set participate in facet count calculation for the top 20 values displayed for CDI.
Maximum Number of Displayed Facets – This field specifies the maximum number of facets that can be displayed in the FE (for example, when More is invoked).
Full Facets Cache – This field indicates whether cache is used to increase the maximum value of the Number of Top Hits for Facet Creation field to 5000. For more information on adding additional memory, see Full Facets Cache (on-premises installations only). For multi-tenant customers, this value may be set as high as 10,000."
"By default, facets in Primo VE return up to 20 facet values for local searches and up to 20 more facet values for CDI in blended searches. This parameter allows you to increase the maximum number of facet values to 50 for local searches only. The valid values are 20–50. CDI will continue to display a maximum of 20 values in blended searches." by discovery_facet_limit
Thank you for clarifying, Stacey
I understand that parity issues have an additional weight.
Stacey van Groll commented
That is a fair point, Manu, I was attempting here to scope a hopefully reasonable request to restore what we once enjoyed in PCI, working on the information provided by Ex Libris in my case stating that this will be an enhancement request for specifically a change in CDI behaviour due to the newly imposed limits.
I did not have many scenarios in a PCI blended where the number of returned facet values for dynamic facets seemed lacking, which is very much not the case on CDI with the much lower limit. I recognise in this as well that there are real system-based limits, but I would expect for a new mandatory system that all factors should be an improvement rather than a downgrade.
Why only CDI and not all facets?