ThoughtSpot is a next-generation BI tool that addresses a specific pain point that almost all data practitioners have encountered — by the time an analytics solution is ready for me and my team to use, the business needs have changed. Can’t we just get direct access to the data so we can ask our own questions?
The answer is the natural language query experience, where we connect to the data and start asking it questions. To optimize the search experience, we currently can create Worksheets to provide a simplified snippet of data for users to search, i.e. filter only relevant columns of data, set default aggregations, specify formatting, standardize formulas, or configure user access.
However, Worksheets are a one-dimensional solution that cannot cope with complex data modeling requirements; more granular tasks have to be pushed back to the data warehouse. The good news is that, as of version 9.10.0., ThoughtSpot introduces Models to optimize the way we work with data connections, and Worksheets will eventually be deprecated.
Deep Dive into the New Models
We can create Models from the Data tab, where we’re also given a choice of cloud data, dbt models, TML (ThoughtSpot Modeling Language) and SpotApps templates. In this case, I’ve got an existing Snowflake connection with a number of tables that I’m going to create my model from:
Once you’ve set up the connection, you can start building your model by simply dragging and dropping tables onto the available canvas. Select the join type (Inner, Full Outer, Left Outer, Right Outer), key field and cardinality. Be sure to add columns from the Columns tab, otherwise you won’t be able to save changes to the model. As you can see in the screenshot below, you can set up formulas, specific filters, parameters and other settings in the same window:
Adding Columns on Top
Next, we move to the Columns tab, where we can either drag the table name onto the canvas to add all columns or select individual columns we want to search on. Note also that repeated column names, such as the key fields, are highlighted in red, so you can remove duplicates as well as set data types and aggregations. Other tabs allow you to add custom formulas, add data source filters and add parameters. Give your model a title and description and click Save.
And that’s it — once you’ve selected your columns, you can start exploring your data. Before you do, consider adjusting your column metadata to make it more search friendly. The Model Editor lets us build our data models visually, and the tabbed interface lets us quickly switch between views:
The last major addition to Models is the bulk edit mode for columns, which previously had to be done in TML (ThoughtSpot Modelling Language). Select multiple columns, click the Edit drop-down menu and adjust the settings.
ThoughtSpot is committed to innovation, and these new feature models are just the beginning. And if you’re just starting out or need more technical knowledge, check out our other blogs on search analytics or contact us for help!