Summary

Product models naturally grow more complex over time. The Product Catalog Analyzer using AI helps product administrators find structural catalog issues through property-graph analysis and natural-language exploration, so they can focus remediation on the model areas most likely to affect quoting and order capture performance.

Large Siebel catalogs can be difficult to investigate because the relevant evidence is distributed across products, classes, relationships, domains, constraints, and promotions. A manual review is time-consuming, incomplete, and hard to repeat as the catalog changes.

The Product Catalog Analyzer using AI is an approach for deeper, cross-entity catalog analysis. It moves selected catalog data from Oracle Database 19c to an Oracle Database 26ai property graph. Catalog entities are represented as connected vertices and edges, enabling multi-hop relationship traversal and impact analysis outside the live Siebel application workload.

The AI model is trained on a Siebel catalog ontology: it understands Siebel classes, attributes, products, product promotions, and versioned-object concepts. This context helps product administrators express catalog questions in familiar business language and interpret findings in the context of their product model.

When to use it - Use the graph-based approach when a catalog has outgrown focused rule-based checks, or when teams need broad, connected evidence before deciding which product-model issues to correct first.

Key benefits

• Use a model trained for Siebel entities. The AI model understands the Siebel ontology, including classes, attributes, products, product promotions, and versioned objects, so product administrators can ask questions in familiar catalog terms.
• Find model hot spots faster. Ask natural-language questions to identify patterns such as large relationship domains, missing target classes, inactive products, or duplicated constraint rules.
• Trace connected catalog behavior. Follow relationships across products, classes, domains, constraints, and promotions rather than reviewing each object in isolation.
• Protect the live application. Perform ongoing analysis in Oracle Database 26ai after migration, reducing query load on the live Siebel environment.
• Make analysis repeatable. Refresh data and rerun the same graph queries to validate remediation and measure change over time.
• Support incremental refreshes. Use a date cutoff to update the graph with changed catalog data instead of recreating the entire data set.
• Enable extensible AI-assisted exploration. An MCP server exposes graph-query tools for conversational analysis through an MCP client. Product administrators can extend the supported analysis with additional natural-language use cases as catalog governance needs evolve.

How it works

The implementation combines a data-migration module, a property graph definition generator, and an MCP server for AI-assisted querying. It requires Python, Oracle connectivity and Instant Client, an Oracle Database 26ai instance with wallet setup, and an MCP client. Because the graph reflects the latest migration, it is best suited to scheduled analysis rather than real-time investigation.

• Select the scope. Identify the catalog entities to analyze. Optionally set QUERY_DATE to limit an incremental extract.
• Migrate catalog data. Run migration_19c_to_26ai.py to stream relational catalog data from Oracle Database 19c to Oracle Database 26ai in configurable batches. The migration uses idempotent upserts, allowing refreshes to be rerun safely.
• Create the graph. Run create_property_graph.py to read the graph-model JSON and define the vertices, edges, key columns, and labels in the property graph.
• Expose analysis tools. Run property_graph_mcp.py to make graph data, vertices, and edges available through MCP tools..
• Investigate and remediate. Use PGQL or natural-language prompts to traverse relationships, identify the specific objects to correct, then rerun the analysis to verify the result.

Supported analysis use cases

The sample prompts cover the following catalog checks. Product administrators can adapt their thresholds and product names, and extend the analysis with additional natural-language use cases to meet evolving catalog-governance requirements.

• 1) Domain count by classCount distinct domain products per Class or Dynamic Class relationship.
• 2) Domain count by productFind product relationships whose domain count exceeds a selected threshold.
• 3) Classes with no target classFind Class or Dynamic Class relationships with an empty Sub Object Class.
• 4) Single-entry domainsFind products with a Class or Dynamic Class relationship containing exactly one domain entry.
• 5) Mixed product and class relationshipsFind Product-type relationships where both a product and a class are populated.
• 6) Products with no target classFind products with Class or Dynamic Class relationships missing a target class.
• 7) Duplicate active constraint rulesFind latest-version products whose active rules share the same rule specification.
• 8) Products including one productFind latest-version products that include a named product directly or through a relationship domain.
• 9) Products including a listFind products that include selected product names and return the results in a table.
• 10) Inactive child productsFind latest-version products containing inactive child products, including direct and domain paths.
• 11) Inactive products in promotionsCheck whether inactive products or their ancestors are included in active promotions.
• 12) Inactive product inventoryList latest-version products whose product definition or current VOD version is inactive.

Accessing the toolkit

Request the Property Graph-Based Product Analyzer Toolkit through an Oracle Support Service Request for Siebel CRM and reference Patch ID 39161636. The package includes a ReadMe, product-analyzer demo video, Python migration, property-graph, and MCP server scripts, a graph-model definition, a configuration template, dependency requirements, and rules guidance for safe PGQL query generation.

Conclusion

The Product Catalog Analyzer using AI provides product administrators with a repeatable analysis layer for large and connected Siebel catalogs. Teams can start with a focused diagnostic baseline, then use graph-based analysis where they need deeper evidence about relationship complexity and its operational impact.

Use the results as part of a continuous remediation cycle: triage high-impact findings, correct the catalog model, refresh the graph, rerun the analysis, and measure improvement. This turns catalog health analysis from a one-time investigation into a maintainable operational practice.

For a technical overview refer to the article: How to Analyze Large Siebel Catalogs with the Property Graph Toolkit.