Why questions in Spotter

When you want to investigate further on a chart in Spotter, you can now ask direct questions like, "Why did my sales drop last month?" or "Why did sales drop in Q2?" Spotter identifies the key attributes and leverages the change analysis engine to provide a clear, natural language summary identifying the key drivers, supported by detailed visualizations. You can easily investigate changes in your data, moving from surface-level observations to insights in the same conversation.

Enabling this feature allows Spotter to send query-related data, such as aggregated metrics and dimension values, to an LLM to identify the data points and generate explanations. Please ensure this complies with your organization’s policies. The data is not utilized for model training, and we maintain a zero-data retention policy with all our LLM providers.
To enable this feature, contact your administrator. For more information on how to enable Why questions in Spotter, see AI feature enablement.

Ask a why question

Once you have asked a business question like, “What’s my monthly sales?” or have an existing visualization, you can ask a natural question starting with “Why”. Spotter will use change analysis to answer the Why questions about a single metric on simple, non-sliced charts. Below are a few examples:

Why has the revenue dropped?

Identifies the revenue metric, picks up the last two date points to identify the key change drivers.

Explain the change in website traffic last week compared to the week before.

Compares the website traffic metric for this week versus the previous week.

Why are churn rates higher in East than West?

Compares the churn metric in East and West.

Why question process

Here’s an example for asking why questions in Spotter:

  1. Start with a visualization: Begin with a chart in Spotter that shows a metric you’re interested in, such as your monthly sales. Or launch Spotter from a visualization in a Liveboard that has a metric.

  2. Ask why: When you notice a change, like a drop or a spike, simply ask a natural language question in the search bar. For example, after seeing a chart of monthly sales, you could ask, “Why did sales drop last month?”

  3. Review the analysis: Spotter performs change analysis and presents the findings in a structured manner:

    • Analysis plan: Spotter first explains how it understands your question and which attributes it will analyze to find the answer.

      spotter why month

    • Detailed visualizations: Spotter then provides detailed change analysis charts that break down the contributions from different attributes, giving you the full evidence behind the findings.

      spotter answer why

    • Narrative summary: Finally, Spotter provides a concise, natural language summary of the key drivers so you can quickly understand the main takeaways.

      spotter narrative summary

Limitations

Why questions in Spotter have the following limitations:

Chart types

This feature works best on charts with a single measure, and no slicing by color or other attributes. It may not provide an answer for multi-line or complex stacked charts.

Ambiguous questions

If your question is ambiguous (for example, you ask about “March” when multiple years of March data exist), Spotter will return an error and prompt for clarification. For ambiguous timeframes like “why did sales drop?”, Spotter will default to comparing the last two data points and state this in its analysis.

Data granularity

This feature may not work as expected if the x-axis has a very high number of data points (for example, >1000).

Was this page helpful?
×