You’re encountering a multi-layered filtering problem that’s particularly common after 23C upgrades. I’ll address the dashboard prompt mapping, subject area joins, and BI cache issues systematically.
First, dashboard prompt mapping verification: The 23B to 23C upgrade introduced stricter validation for dashboard prompt variables. Navigate to your dashboard in edit mode, click on the product category prompt, and examine the ‘Prompt Properties’. Note the exact variable name - it should be something like ‘pv_ProductCategory’. Now open each embedded analysis (historical sales and forecast) separately in edit mode. Click ‘Edit’ > ‘Prompt’ and verify that the prompt variable name EXACTLY matches the dashboard-level variable, including case sensitivity. In 23C, Oracle changed from case-insensitive to case-sensitive variable matching. If your dashboard uses ‘pv_ProductCategory’ but your forecast analysis uses ‘pv_productcategory’, the filter won’t apply. Standardize all variable names to match exactly.
Second, subject area joins and column selection: Your historical analysis uses ‘Sales History - Product Hierarchy’ while forecast uses ‘Demand Planning - Forecasts’. These are separate subject areas that don’t automatically share filter context unless you use conformed dimensions correctly. Here’s the fix: Edit your forecast analysis and remove the subject-area-specific product category column. Instead, add the column from the conformed dimension ‘Items - Product Hierarchy.Product Category’. This dimension is shared across both subject areas. Do the same for your historical analysis. When both analyses reference the SAME dimensional column path (not just similar column names), the dashboard prompt applies consistently. You can verify the column path by hovering over the column in the criteria pane - it should show the full path starting with the dimension name.
Third, BI cache configuration: The 23C upgrade reset some cache policies to more aggressive defaults. Navigate to Analytics > Administration > Manage Cache Policies. Search for policies related to ‘Demand Planning’ subject areas. You’ll likely find that the cache persistence time was increased from 1 hour to 4 hours during upgrade. For dashboards with dynamic filters, this causes stale data issues. Reduce the cache persistence to 30 minutes or less for your demand planning analyses. Additionally, enable ‘Refresh Cache on Query’ for these specific analyses. To do this, edit each analysis, go to Administration tab, and check ‘Disable Cache’ during your testing phase. Once filters work correctly, re-enable cache with reduced persistence time.
Fourth, cross-subject-area filter propagation: In 23C, Oracle introduced a new setting that controls how filters propagate across subject areas in combined dashboards. Navigate to Analytics > Administration > Manage Repository Variables. Look for a variable named ‘CROSS_SUBJECT_FILTER_ENABLED’ or similar. It should be set to ‘true’. If it’s ‘false’ or doesn’t exist, create it as a Static Repository Variable with value ‘true’. This ensures that when a dashboard prompt filters on a conformed dimension, the filter applies to all subject areas that include that dimension, even if they’re not directly joined in the physical model.
Fifth, dashboard-level configuration: Edit your dashboard and click ‘Dashboard Properties’. In the ‘Filters and Variables’ section, ensure ‘Apply Saved Filters on Load’ is enabled. Also check ‘Propagate Filters to All Pages’. In 23C, these settings default to disabled for dashboards migrated from earlier versions. Without them, filters may apply to the first page only or not persist across user sessions.
Sixth, analysis-level filter configuration: Open your forecast analysis in edit mode. In the Criteria tab, click on the product category filter. In the filter properties, verify that ‘Is Prompted’ is checked and the prompt matches your dashboard variable name. Critically, also check ‘Use Values from Prompt’ rather than ‘Specify Values’. Some upgrades reset this to ‘Specify Values’ which causes the analysis to ignore dashboard prompts and use its own static filter values instead.
Finally, troubleshooting with the BI Diagnostics tool: If filters still don’t work after the above fixes, enable detailed logging. Go to Analytics > Administration > Manage Diagnostic Logs, set ‘Query Engine’ to level 5 (verbose). Run your dashboard with a filter applied, then examine the query log. Search for your forecast analysis name. You’ll see the actual SQL generated. Look for WHERE clauses - if the product category filter is missing from the SQL, it confirms the prompt isn’t being applied. The log will show the exact point where filter propagation fails, helping you identify if it’s a prompt mapping issue, subject area join issue, or something else.
Given that this broke after 23B to 23C upgrade, I’d also recommend checking Oracle Support document 2891456.1 which covers known issues with dashboard filtering in 23C. There’s a specific bug (Bug 35789234) where dashboard prompts don’t cascade to analyses using certain demand planning subject areas. If your configuration matches that bug pattern, you may need to apply a patch or use a workaround involving combined analyses rather than separate embedded analyses.