This session detailed how InterSystems IRIS addresses the critical need for a capable platform to manage and leverage complex data, especially in the context of rapidly moving data supply chains. The core message revolved around transitioning from an operational model to one that emphasizes data visualization, reporting, and machine learning, all facilitated by InterSystems IRIS and its complementary tools.
The Data Supply Chain and Analytical Model
The concept of a “data supply chain” was introduced using a FedEx example: just as a package undergoes various hand-offs and unifications before reaching its final destination, data also moves through multiple points of relocation or transformation. The end-user, however, simply sees the package arrive where they need it. This process mirrors the analytical data model, which aims to simplify the user interface and queries. The goal is to make data digestible for end-users, much like how Excel presents one table at a time for ease of consumption. InterSystems emphasizes the importance of a capable platform for this complex data movement.
Scenario 1: Application Development and BI
Designing effective analytical models begins with understanding the end-user’s requirements and the available data. The process involves flattening and de-normalizing data to support diverse use cases. InterSystems IRIS BI is specifically designed to transform transactional data into analytical models. Key questions guiding this process include: What information is most vital to track? How should it be summarized and visualized? How will it be used, stored, and displayed?
Application developers organize vast amounts of operational data within the UI’s central pane, establishing relationships and drill-downs. The next step involves viewing this organized data in the Analyzer tool, which allows users to drag and drop elements to create tables and provides click-through drill capabilities for on-the-fly data summarization. InterSystems IRIS BI also offers existing Dashboards that are embeddable in applications, allowing for customization through filtering and sorting. These dashboards are intended for end-users, not specialized analysts.
The typical workflow involves an architect flattening information and identifying the most crucial data for the target audience. This refined data is then handed off to an analyst, who uses the Analyzer for drag-and-drop exploration. Ultimately, end-users interact with the curated dashboards.
Scenario 2: Operational Modeling and Machine Learning
This scenario focuses on predictive analytics, exemplified by the prediction of delivery delays. The process involves connecting to InterSystems IRIS, fetching data, performing quality checks, and then using an H2O model for machine learning. In the example used by the presenters, the model generates predictions and returns metrics, along with visualizations comparing actual versus predicted delays. Crucially, the trained model can be saved to the database and integrated into subsequent work, such as future dashboards or an analyst’s Excel sheets. The Notebook environment is the preferred tool for performing machine learning tasks.
Scenario 3: Consumer Analytics and Executive Reporting
This scenario explores how executives consume data. Users can open Excel and pull data directly via an ODBC connector, examining the SQL rendering of a cube definition. This process allows analysts to manipulate data at a granular level using previously assigned column names.
Adaptive Analytics and Natural Language Querying
Adaptive Analytics is presented as an add-on product that elevates analytical modeling. It provides a rich dimensional modeling experience, feeding into popular consumer BI tools like Power BI, Qlik, Excel, and Tableau. A key feature is its intelligent aggregation and partial result building, which optimizes common query patterns for faster performance.
The session also introduces Natural Language Querying (NLQ) with InterSystems IRIS, enabling users to transform natural language questions into SQL queries. This leverages Generative AI to create, run, and interpret the results of these queries. The process begins with a user question and involves using embeddings for specific column data, embeddings for the entire table structure, and the table structure in DDL (Data Definition Language).
A detailed example was provided: for the question “Which suppliers can deliver syringes before July 4th?” the system identifies “July 4” as a date and “syringes” as a product type. It then identifies relevant tables and uses schema information to determine how these tables should be connected. This structured information is then presented to an LLM (Large Language Model), which is prompted to generate an SQL statement to answer the query. The system then runs the query and provides an explanation of the results.
Unified Data and Future Vision
Despite these diverse scenarios, the session emphasizes that all rely on the “same data” and leverage the same data supply chain. InterSystems IRIS BI was created for designing reusable analytical models, aiming to help organizations organize data within a single platform, from operational data to universal analytical consumption.
Roadmap
Presenters shared the following future offerings:
- IRIS BI Boost: A new rendering layer for existing BI Dashboards.
- Flexible Data Materialization Options: To extend IRIS BI models beyond MDX cubes.
- Notebook Experience: Quick spin-up from relevant dashboards and query contexts.
- Templates for RAG and Natural Language Query Environment: To streamline development.
- AI Gateway: To accelerate augmented application development.
Benefits of InterSystems IRIS
The benefits of InterSystems IRIS were highlighted, including:
- Unified Storage and High-Performance Querying: For all data types (relational, document, object, vendor, etc.) with common metadata.
- “Shift Left” Analytics: Data remains within IRIS, promoting model reuse.
- Data Fabric Studio: For building pipelines to ingest and organize data within IRIS.
- No- and Low-Code Interfaces: Especially for FHIR resources.
Vision for Analytics and Data Management Maturity
The vision for analytics centers on the concept of a “data product,” defined as an integrated, curated, and self-contained combination of data, metadata, and semantics. The goal is to move beyond a producer-centered data catalog (the current state) to a consumer-centric data marketplace. In this future state, data quality is guaranteed, published metrics are available, users can subscribe to updates, and data is governed, ensuring it is “fit for use.”
To achieve this vision, ongoing work includes data discovery in source systems, profiling, data quality assessments, and augmentation. Emphasis is placed on making metadata comprehensive to ensure high-quality data products and establishing a fully consumer-focused marketplace for an entire data product portfolio.
Q&A
In the Q&A portion of the session, the presenters confirmed that InterSystems IRIS BI is included with most IRIS licensing options. A new dashboarding look will be available simply by accessing a new URL and will be covered in a different session at this year’s conference. Early Access Program participation is encouraged.
