Differences between Snowflake Connected and Native App
QPR ProcessAnalyzer with Snowflake calculation is available in two deployment options: Connected App and Native App. Both run process mining analyses on data stored in Snowflake, but they differ in where the application runs, how it is accessed, and which features are available.
Overview
In the Snowflake Connected App, QPR ProcessAnalyzer runs outside Snowflake and connects to a Snowflake account through an ODBC connection. QPR ProcessAnalyzer Server generates the SQL queries needed for the calculations and sends them to Snowflake for execution, while the eventlog data stays in Snowflake.
The Snowflake Connected App is available in two variants:
- Managed: QPR Software provides and runs QPR ProcessAnalyzer as a managed service, and the customer connects it to their Snowflake account.
- Self-hosted: The customer installs and runs QPR ProcessAnalyzer in their own infrastructure (cloud or on-premises) and is responsible for operating, updating and maintaining it.
In both variants, the application runs separately from Snowflake and is accessed with a web browser through its own web address.
The Snowflake Native App runs entirely inside the customer's own Snowflake account as a container-based application using Snowpark Container Services. It is installed from the Snowflake Marketplace. All data stays in the customer's Snowflake account, and the app uses Snowflake's security, access controls and governance functionality. The Native App is updated automatically, so users do not need to perform any update actions.
Deployment options comparison
| Aspect | Snowflake Connected App | Snowflake Native App |
|---|---|---|
| Where the application runs | Outside Snowflake, either on the customer's own servers (self-hosted) or as a managed service in QPR Cloud provided by QPR Software. | Inside the customer's Snowflake account, as a Snowpark Container Services application. |
| Accessing the app | Through the QPR ProcessAnalyzer web address using a web browser. | Opened from the application page in Snowsight. Access is granted by assigning the USER application role to Snowflake account roles. |
| User management and authentication | Users authenticate with QPR ProcessAnalyzer or Snowflake user credentials. SAML 2.0 authentication is also available for connecting to an external identity provider. | Users authenticate with Snowflake user credentials. |
| Access to Snowflake data | The connection string contains a single Snowflake user identity that determines data permissions for all users. Alternatively, each user can access Snowflake data with their own Snowflake user permissions. | The Native App uses one common application identity to manage data access for all users. |
| AI Assistant | AI Assistant is available. | AI Assistant is not available. |
| AI Root Causes Analysis | AI Root Causes Analysis is available. | AI Root Causes Analysis is not available due to current Snowflake limitations. The classic root causes analysis is still available. |
| MCP server | QPR ProcessAnalyzer can be configured as an MCP server that agents (such as Claude, Microsoft Copilot and OpenAI GPT) can connect to using OAuth or an API key. | The MCP server is always available, and Snowflake Cortex Agents and CoWork can access it. It can also be accessed from outside Snowflake using programmatic access tokens (PATs), but OAuth is not supported for external access. |
| Connecting to external data sources | Available, for example, for SAP, Salesforce, ODBC and web service sources. | Not available in QPR ProcessAnalyzer. Data must first be loaded into Snowflake using Snowflake features, such as Snowpipe or Openflow. |
| Scripting | Both expression language and SQL scripts are available. | Expression language scripts are available, but SQL scripts are not. SQL operations can be run using Snowflake procedures. |
| Installation |
Self-hosted: The customer installs the software and configures the Snowflake connection. |
Installed from a Snowflake Marketplace listing or a private listing. |
| Updates |
Self-hosted: The customer's administrator updates the environment. |
QPR Software releases updates, and Snowflake automatically applies them to installed Native Apps. |
| Data storage | Data can be stored and managed in Snowflake and SQL Server tables. | Data can be stored and managed only in Snowflake tables. SQL Server storage is not available. |
| Data caching | One schema in the Snowflake account is needed where QPR ProcessAnalyzer can create temporary cache tables. Write permissions to that schema is needed. | Native App uses its own database write temporary cache tables. No further permissions need to be given for this schema. |
| Logging and monitoring | QPR ProcessAnalyzer writes a log file that is available to administrators. | Logs, traces and metrics are written to the Snowflake event table and can be viewed in Snowsight. |
| Configuration data backups | Handled as part of the self-hosted installation, or by QPR Software in the managed service. | Snapshots of the container block storage are created and restored using Snowflake SQL commands. |
Common features of both deployment options
Both options use the same QPR ProcessAnalyzer product with Snowflake, so the following applies to both:
- Eventlog data is stored in Snowflake tables, and the process mining data models are mapped to these tables. Readonly access is suffient to the eventlog data, if the intention is not to ingest data using QPR ProcessAnalyzer.
- Process mining calculations are run in Snowflake using the virtual warehouses. All types of warehouse configurations can be used to optimize performance (except the interactive warehouses are not optimal for the process mining analyses).
- All data stays in the customer's Snowflake environment and is not moved or duplicated outside Snowflake. In the Connected App, calculation results shown in dashboards are loaded outside Snowflake (for showing them to users) but they are not stored permanently outside Snowflake.
- Snowflake's security, access controls and governance apply to the data in Snowflake.
- QPR ProcessAnalyzer offers an MCP server interface that gives AI agents access to process intelligence.