How to Set Up ThoughtSpot Spotter for AI Analytics

What ThoughtSpot Spotter Actually Does

ThoughtSpot Spotter is a natural language query engine built into ThoughtSpot's analytics platform. You type a question in plain English, and Spotter translates it into a query against your connected data warehouse. It returns charts, tables, and summaries without requiring you to write SQL or build dashboards manually.

Spotter works on top of semantic models you define in ThoughtSpot. These models map your raw database tables into business-friendly terms. When you ask "what were total sales last quarter by region," Spotter uses that model to generate the correct query, apply the right filters, and return a visualization.

This matters for ops managers, analysts, and founders who need answers from their data but do not want to wait for an engineer to build a dashboard. According to Vendr data, the average ThoughtSpot enterprise contract runs about $137,000 per year. But the free trial and the $25/user/month Essentials plan make it accessible for smaller teams to test first.

Step 1: Sign Up for a Free Trial

Go to thoughtspot.com/trial and enter your business email. ThoughtSpot will send an activation link. Click it to verify your email and access your trial instance.

The free trial runs for 30 days and includes sample retail data preloaded so you can start querying immediately. You can invite up to five teammates from the Admin tab.

Once logged in, you land on the home screen. The Spotter interface is accessible from the app switcher menu in the top left corner.

Step 2: Connect Your Data Source

If you want to analyze your own data instead of the sample set, you need to connect a cloud data warehouse. ThoughtSpot supports Snowflake, Amazon Redshift, Google BigQuery, Databricks, Azure Synapse, and Starburst.

To add a connection, go to the Data workspace, select Connections, and click "Add connection." Choose your warehouse type and enter the connection credentials: host, port, database name, and authentication details. ThoughtSpot will validate the connection and show available schemas and tables.

For your first data source on a Spotter-specific plan, ThoughtSpot may require a brief onboarding call to help configure the initial connection. After that, you can add more connections on your own through the same Data workspace.

Step 3: Build a Semantic Model

A semantic model is the layer between your raw tables and the questions users ask. It defines which columns are dimensions, which are measures, how tables join, and what business-friendly names to use.

In the Data workspace, click "Create Model." Select the tables you want to include from your connected data source. ThoughtSpot will auto-detect column types and suggest joins based on matching column names.

Review and adjust the following:

Column names: rename technical column names to plain language. Change "rev_amt_usd" to "Revenue (USD)" so Spotter understands what users mean when they ask about revenue.

Relationships: confirm that join paths between tables are correct. If you have a sales table and a regions table, make sure the foreign key relationship is properly mapped.

Aggregation defaults: set default aggregations for measure columns. Revenue should default to SUM. Customer count should default to COUNT DISTINCT.

Save your model. This step typically takes 15 to 30 minutes for a straightforward dataset with 3 to 5 tables.

Step 4: Add Natural Language Instructions

This is what separates a basic setup from one that gives accurate answers. ThoughtSpot lets you add plain-text instructions to your model that guide how Spotter interprets queries.

Open your model in the Data workspace, click the Instructions tab, and type rules. For example:

"When a query does not specify a time period, default to the last 30 days."

"When calculating revenue, always exclude transactions where Account_Type equals Internal Test."

"For queries about Electronics sales, filter for Fulfillment_Center equals FC-West."

Write these instructions using exact column names and exact values from your data. Vague instructions like "usually exclude test data" will not work reliably. Be specific and imperative.

After saving, test each instruction by running the relevant query and clicking "Regenerate last answer" to verify Spotter follows the rule.

Step 5: Enable Spotter on Your Model

Not every model has Spotter enabled by default. To turn it on, navigate to the Data workspace, select your model, click the three-dot menu in the upper right corner, and select "Enable Spotter."

Once enabled, users can open the Spotter interface and select your model as their data source. The model name appears in the data source picker at the top of the Spotter chat.

Step 6: Ask Your First Question

Open Spotter from the app switcher. Select your model from the data source dropdown. Type a question like "show me monthly revenue trend for the last 12 months."

Spotter will generate a visualization. You can follow up with refinements: "break that down by region" or "only show the top 5 regions by total." Each follow-up builds on the previous context, so you can drill into data conversationally.

If the result looks wrong, check your model definitions and instructions. The most common issues are missing joins, incorrect aggregation defaults, or ambiguous column names.

Step 7: Share and Schedule

Once you have a useful answer, you can pin it to a Liveboard (ThoughtSpot's term for a dashboard), share it with teammates via link, or schedule it as a recurring email report.

Click "Pin to Liveboard" on any Spotter result to save it. From the Liveboard, click the schedule icon to set up automated delivery. Reports can be sent daily, weekly, or monthly as PDF or CSV attachments.

Troubleshooting Common Issues

The most frequent problem new users hit is Spotter returning unexpected results. This almost always traces back to the semantic model. Check these areas first:

Ambiguous column names cause Spotter to guess which column you mean. If you have "date" in three different tables, rename them to "order_date," "ship_date," and "return_date" in your model.

Missing joins mean Spotter cannot connect related data. If you ask about "revenue by customer segment" but your revenue table and customer table are not joined in the model, the query will fail or return incorrect totals.

Stale cache can also cause problems. After making model changes, allow a few minutes for the cache to refresh, or manually clear it from the Data workspace settings.

If Spotter consistently misinterprets a specific type of question, add a targeted instruction in the Instructions tab. One well-written instruction is more effective than five vague ones.

What to Know About Pricing

ThoughtSpot's free trial gives you 30 days with sample data and your own connections. After that, the Essentials plan starts at $25 per user per month for 5 to 50 users with up to 25 million rows. The Pro plan at $50 per user per month adds agentic AI features and more capacity. Enterprise pricing is custom and typically starts around $100,000 annually.

ThoughtSpot also charges based on query volume through a credits system. Every search, dashboard refresh, and background query consumes credits. Monitor your usage in the Admin console to avoid unexpected overages.

If your team is smaller than five people and your data fits in a CSV, tools like VSLZ handle the same natural language analytics workflow from a file upload with no warehouse connection or model setup required.

Practical Summary

Start with the free trial and sample data to learn the interface. Connect your warehouse and build a model with clear column names and join paths. Add specific natural language instructions to improve accuracy. Enable Spotter on the model and test with real questions. Pin useful results to Liveboards and schedule recurring reports. Track your credit usage from day one to forecast costs accurately.