Getting the Best Results with Minds

Minds translates natural language into SQL queries against your data. Think of it as a data analyst on your team — it's capable, but needs clear context to give you the right answer.

Be Specific

Every good question includes three things:

• A metric — revenue, customer count, profit margin
• A dimension — product, region, customer type
• A timeframe — Q2 2024, January 2025, last 12 months

Vague	Specific
"How are sales doing?"	"What was total revenue by region in Q2 2024?"
"Was customer churn high?"	"What was the churn rate for enterprise customers in Q4 2024?"
"Which product sold best?"	"Which product had the highest total units sold in Q3 2024?"
"How did Product A perform?"	"What was Product A's revenue and growth rate in Q1 2024?"

Break Complex Questions Into Steps

If a question involves multiple analyses, ask it as a conversation rather than a single sentence.

Instead of: "Which region had the highest revenue growth in 2024, and how does that compare to the top-performing product?"

Try:

1. "What was revenue growth by region in 2024?"
2. "Which region had the highest growth?"
3. "Now compare that to the top-performing product's growth."

This also works for medium-complexity queries:

• "Compare churn between Q1 and Q3 2024" becomes:
  1. "What was the churn rate each month in Q1 2024?"
  2. "Now show Q3."
  3. "Compare the two."

You can drill down conversationally at any point — Minds keeps context across messages:

• "Break that down by customer segment."
• "Now show me Q3 only."
• "Which region contributed most?"

Configure Your System Prompt

Use the System Prompt settings to give Minds standing context about your business and preferences.

Business context — define terms, acronyms, and metrics specific to your organization:

• "Our fiscal year starts in April."
• "NRR means net revenue retention."
• "Use 'active customer' to mean anyone with a login in the last 90 days."

Response style — control how Minds communicates:

Goal	System Prompt
Fast, factual answers	"Be concise. Show the data, skip interpretations."
Insightful analysis	"Explain what the numbers mean and highlight trends."
Strategic input	"Propose ideas or next steps based on the results."
Risk-aware reasoning	"Be cautious, only share conclusions with strong evidence."

Know the Strengths and Limits

Works well:

• Filtering by date, product, or region
• Aggregations — count, sum, average, growth rate
• Comparisons across time periods or categories
• Top-N rankings

May need adjustment:

• Listing large result sets (e.g. "list all customers") — ask Minds for the SQL and run it directly
• Multiple conditions or time periods in one sentence — split into steps
• Questions using derived metrics that haven't been defined — add them to your system prompt
• Missing key filters (product, region, timeframe) — be explicit

Quick Reference

Query Complexity	Approach
Simple — single metric, single filter	Ask directly: "What was revenue in EMEA in Q2 2024?"
Medium — aggregation or trend	Ask for the base data first, then compute: "Total revenue by region?" then "Which was highest?"
Complex — comparison across dimensions	Decompose into steps: get each piece separately, then compare