Spotter 3 best practices

This page describes how to use Spotter 3 effectively.

Overview

Spotter 3 introduces significant improvements over previous versions:

  • Autonomous analysis — Translates natural language to data queries, plans analysis, and delivers insights

  • Smarter context — Detects new questions vs. follow-ups automatically

  • Clarification requests — Asks questions when uncertain instead of guessing

  • Transparent reasoning — Shows analysis steps in plain language

  • Narrative summaries — Provides insights, not just charts

Ask effective questions

Be specific about your intent

Spotter handles open-ended questions but uses its own judgment to plan analysis. Specify dimensions explicitly if needed.

Less effective More effective

"How are sales doing?"

"How are sales doing? I expect analysis at region and store level."

"Show me customer data"

"Show me customer acquisition by channel."

"What’s our performance?"

"Revenue performance this quarter. Compare it against last quarter and same quarter last year."

Use Research Mode for complex analysis

  1. Ask a high-level question (for example, "Analyze customer retention trends and identify risk factors").

  2. Review the proposed analysis plan.

  3. Add or remove steps as needed.

  4. Approve the plan.

  5. Receive a comprehensive report.

Research Mode works best for quarterly reviews, root cause analysis, and multi-dimensional investigations.

Use Auto-mode

Use Auto-mode if you don’t know which model to use. Auto Mode also supports cross-source queries in case you want to query data from multiple sources to generate insights:

"Show high-usage customers and their open support tickets."
"Which sales reps have the most deals but lowest close rate?"
Manual data model selection improves accuracy when models have overlapping data.

Brainstorm with Spotter

Provide role context to get analysis suggestions:

"I'm a PM focused on engagement. What metrics should I track?"
"I manage sales. What questions should I ask about our pipeline?"

Verify and use results

Review the reasoning

  1. Expand the reasoning block in the response.

  2. Check how Spotter interpreted your question.

  3. Verify the data sources used.

  4. Review key decisions made during analysis.

Unsure about a response

  1. Ask Spotter to explain its analysis.

  2. Ask Spotter to verify its analysis or double-check a specific result.

Request clarification

If results seem incorrect, ask Spotter to explain:

"How did you calculate that percentage?"
"Why did you exclude those records?"
"Double-check this calculation"

Identify non-data responses

Spotter can answer open-ended questions, but ambiguous queries may produce responses from general knowledge instead of your data.

Signs of a non-data response: - Generic-sounding answer - No specific numbers from your data - Reasoning block shows no data queries

To fix this

Be more specific, or select a data model before asking.

Feature-specific guidance

Research Mode

Always review the analysis plan before approval. Add specific dimensions or comparisons you need.

  • Expect longer processing time for thorough analysis.

  • Remove irrelevant steps to save time and cost.

  • Don’t close the tab while running Research Mode to ensure Spotter finishes its analysis.

Auto Mode

Works best when the overlap between available data models is low.

  • For overlapping models, specify the data model name if you already know it.

  • Ask Spotter to run queries on multiple models if you cannot choose.

  • Ask Spotter to share more details about suggested data models if you need help deciding.

    For analysts

    End users may have access to fewer Models. Results vary based on user permissions.

Connectors

  • Specify the connector name and how to use it for better results.

  • Add external context using Connectors to bring unstructured content into Spotter.

  • Use Connectors to take actions or share insights from Spotter.

Was this page helpful?
×