Spotter business terms

To begin coaching business terms:

  1. Navigate to the Data workspace.

  2. Under Spotter coaching, select the Business terms tab.

  3. Click on Add business term.

    This will lead you directly into the coaching flow, where you can begin refining responses and improving Spotter’s accuracy. You can use the data source tab next to the Search bar to select the data source for which you want to define a business term.

    Coach Spotter references

  4. In the Enter a business term text box, enter a business term.

  5. In the Search bar, enter the search token(s) that defines the business term for your data, then click Go. For example, you could enter a business term “sales contribution” and define it explicitly as sum(order_amount) / sum(total_revenue). All future questions using “sales contribution” will apply your exact logic.

  6. Your AI-generated Answer appears. Review the search tokens and visualization Spotter chose to visualize your definition and check for accuracy.

    You can modify the search tokens to refine the response. Changing these tokens updates the data visualization, allowing you to better tailor the answer to the question.

    Once you’re satisfied with the adjustments, click Submit to save the response as a reference for future, similar queries.

  7. You return to the Coach Spotter page, where you can see the coaching you’ve added.

Business term feedback level

Depending on your role (for example, Admin or Model editor), saving a business term may apply globally to all users of the Model, or it might be initially saved at user level. For more information, see Understand coaching levels.

Guidelines for effective business terms

Principle Why it matters Example

Meaningful addition

Only define terms that an LLM can’t reliably guess or that have a very specific meaning in your business.

Avoid coaching “customer” as "customer column" if it’s the obvious choice in columns and Spotter is able to pick it up without coaching.

Business-specific logic

Focus on how your company uniquely defines things like revenue, churn, or contribution.

“Sales contribution” might seem obvious but often has a specific calculation method in your business which the LLM might not be able to predict.

Consistency across contexts

Spotter applies business terms globally; the term must hold the same meaning in all contexts it can be used for that Model. So avoid context-specific logic.

Be cautious when coaching general time-related terms if you have multiple date fields. For example, coaching “this year” to always use created_date can cause issues if a user asks about “this year” related to transaction_date.

Correct logic

Ensure the underlying definition (filters, formulas) is accurate and aligns with business intent.

Avoid coaching ‘profit’ as sales - material cost if the definition of profit varies across categories. Coach “material profit” and “net profit” separately.

Similarly, if “best products” means “most selling product” for products in the North American region but “highly-rated products” in the European region, do not save this business term. Instead, coach on two reference questions using this term in different region questions to highlight the difference.

Key considerations

Avoid business terms for obvious model interpretations

Don’t create business terms for simple concepts Spotter infers directly from your data model. For example, if Spotter understands “total sales” as “sum(sales)” from the Model, a business term for “total sales” is unnecessary. Reserve business terms for specific business nuances, complex logic, or synonyms not evident from the Model.

Avoid conflicting definitions

A term must mean the same thing every time it’s used within its defined scope (for example, within a Model). For example, if “annual revenue” is defined as “sum(revenue)” where “period = annual”, don’t try to create another business term “yearly revenue” with a slightly different formula if users will use the terms interchangeably, as this can confuse Spotter.

Don’t create overly rigid coaching for context-dependent terms

For instance, coaching “this year” to always use a very specific, niche date field (for example, “sales date”) can lead to misinterpretations when users ask about “this year” in other contexts. For example, the query, “Number of orders this year?” should use “order date” and not “sales date”.

Was this page helpful?
×