Known issues and limitations
Last updated: 2026-11-19Semantic models in data products have several known limitations that you should be aware of when creating, consuming, and managing them.
Consumption methods
To load data products with their semantic models, you need to open data catalog from one of the following entry points:
-
Open Data load editor in an application, and select Add from data catalog.
-
Open a data product. From its options menu, select Use in > Application.
In other words, you cannot load semantic model information from the data product using entry points such as Data manager or the application creation landing page.
Expression validation
During semantic model consumption, validation is not performed to check if master items are valid. The validation error appears only when you inspect the master item in the app.
IncludeModel script limitations
The WithTables: * wildcard syntax is not supported. You must list tables explicitly.Relationship suggestions
The relationship suggestions feature provides limited support for fuzzy matching.
Master items in consuming applications
You can delete a master item, but it will reappear on the next application reload if it is still referenced in the IncludeModel script. To permanently remove it, remove it from the script first. You can also duplicate any semantic model master item. The duplicate will be treated as an ordinary application master item and will be fully editable.
Import, export, and cross-product reuse
- Exporting a semantic model to a file is not supported.
- Importing a semantic model from another tenant is not supported.
- Duplicating a data product (including its semantic model) is not supported.
Limitations of relationship suggestions from Data Integration pipelines
When you build a semantic model on a data product, Qlik can suggest relationships between datasets based on the relationships defined in the Data Integration data project that produced them. These suggestions are a starting point. They do not always reflect everything modeled in the pipeline, so review them and add any missing relationships manually.
Renamed key columns
A relationship is suggested only when the foreign key can be matched to the primary key of the dimension it points to. If a foreign key is named differently from the primary key of the dimension (for example, customer_ref in a fact table that points to customer_id in a dimension), the mapping cannot be resolved and no suggestion is made for that relationship. Pipelines where keys are renamed along the way will therefore show fewer suggestions than expected.
Physical model, not logical model
Suggestions are currently derived from the relationships of the physical model, not the logical model. The physical model includes technical key columns generated by the pipeline, whose names start with hdr_. These columns can appear in the suggested relationships even though you do not see them in the logical model you designed in Data Integration. Because the logical model is the one you work with in Data Integration, you may expect suggestions to follow it. Relationships that exist only at the logical level are not suggested.