Spotter best practices

The most effective way to get a precise answer is to structure your questions with three key components in mind. Think of it as providing a clear brief to your analyst. A good rule of thumb is to specify the “what”, the “how”, and the “where” you are looking for in your answer.

A good question clearly defines the calculation, grouping, and filters. Here’s how that looks in practice:

You don’t need to ask the perfect, complex question on your first try. Spotter is designed for a conversational flow. Start with a broader question and refine it based on the results.

For example, you might start with a broad question: “What are the sales of the top 5 products by category?” Spotter shows the overall top 5 products and their categories.

Then you can refine your grouping: “I wanted the top 5 products in each category.” Spotter corrects the analysis to show a per-category ranking.

Try adding a filter: “Show data for last week.” The analysis is now filtered to the most recent time period.

Ask a follow-up calculation: “What was the percentage contribution of bags in the total sales last week?” Spotter performs a new calculation based on the context of the conversation.

AI models are powerful, but they shouldn’t be left to guess your intent. Ambiguity is the primary cause of unexpected or inconsistent answers.

Be careful how you phrase your question, to avoid multiple possible interpretations. For example, if you ask “What are the total sales for the Southern region?”, do you mean only the South region, or should the Southwest region be included?

One option is to coach Spotter’s business terms to clarify that South should include the Southwest region. However, if this interpretation is not universally applicable to your business, it is better to be explicit in your questions rather than trying to add coaching to Spotter. You can ask, “What are the total sales for the Southern region, excluding Southwest?” or “What are the total sales for the Southern region, including Southwest?”

You should also be clear about the criteria when asking for rankings. If you ask “What are my most selling products?”, how should Spotter interpret “most selling”? Does it mean highest revenue, greatest quantity sold, or highest profit margin? Instead, ask “What are the top 5 products by revenue for last year?” or “Show me the products with the highest quantity sold last month in the East region.”

To get the right answer quickly and efficiently, always be explicit about time frames and complex calculations.

If you don’t provide a time frame, Spotter will query your entire data history. This can be slow, costly, and rarely provides the insight you actually need. Instead of asking “What are the total sales for the Southern region?”, ask “What are the total sales for the Southern region in Q3?”

If you need a calculation like “percentage contribution,” briefly explain the logic. This ensures Spotter builds the formula exactly as you intend. Instead of asking “What is the percentage contribution of the East region for the last three months?”, say “For the last three months, show me the East region’s sales as a percentage of total sales, and break the results down monthly.”

If you find yourself repeatedly clarifying the same business term (like the Southern region example above), teach Spotter what you mean.

You can click + Add to coaching below an Answer to create custom definitions. If you define the business term “Southern region” as region= ‘South’ OR region = ‘Southwest’, Spotter will use that interpretation in all your questions including that term, saving you time and ensuring consistent analysis across your team.

For more information on coaching, reference Creating reference questions and Defining business terms.

In order to help you understand how Spotter works, we provide a simplified syntax that allows you to validate the output generated by Spotter without learning the SQL syntax.

You can verify Spotter’s results by checking the query and by comparing it to the previous step in the conversation.