Chat with your Data

Ask focused questions, investigate the result and continue earlier conversations.

Updated 9 August 2026

Ask Incendium data a business question

AI → Chat with your Data lets you investigate the project in conversational language. Incendium can look up relevant metrics and dimensions, produce supporting charts, explain the pattern it finds and suggest practical next questions or actions.

The important difference from a general-purpose chatbot is the measurement context underneath it. Well-trained Locations, Channels, Conversions, effectiveness rules and ecommerce data give the analysis a clearer description of the business. Better inputs make both human interpretation and AI-assisted analysis more dependable.

Start in the correct project

Check the client and project selectors before asking a question. Chat uses the data available to the selected project, so a well-written question cannot correct the wrong project context.

Write a question that can be answered

Use the Ask me anything box or select one of the suggested questions. A useful first question normally contains four parts:

Measure

Name the outcome or behaviour: Conversion rate, revenue, ROAS, effective sessions, add-to-cart rate or landing-page engagement.

Scope

Identify the relevant Channel, Campaign, Location, product, device, Audience or funnel stage.

Time

Give a concrete period such as the last 30 days, this quarter or the previous full month.

Comparison

Say what should be compared: devices, Channels, new and returning visitors, Audiences or the preceding period.

For example: Compare Conversion rate and revenue by device for Organic Search over the last 30 days, and show which landing-page Locations explain the largest difference.

Begin with one decision-sized question. If the request contains several unrelated investigations, ask them separately so the result and charts remain easier to verify.

Read the response in layers

While working, Chat shows the analytical steps it is taking. A completed response can contain:

Chart links that open the underlying visual comparison produced for the question.
Insight statements that highlight the most important observed pattern.
An Analysis Summary that connects the measures and explains the likely commercial meaning.
Key Recommendations that turn the analysis into possible next actions.

Use the result as an analytical starting point, not an automatic decision. Open the charts, check the date range, sample size, metric definition and filters, then compare important findings with the relevant Incendium report. A recommendation may be sensible while its assumed explanation still needs testing.

Continue the investigation with follow-up questions

The question box remains available beneath the answer. Follow-ups retain the conversation context, so you can narrow or challenge the result without restating everything.

Ask for the breakdown

Try: Break that difference down by Channel and device.

Check whether it persists

Try: Does the same pattern appear in the previous 30-day period?

Find the journey behind it

Try: Which Locations and Interactions contribute most to this result?

Test the explanation

Try: What evidence supports that recommendation, and what else could explain the change?

If the conversation moves to a different business question, start a new Chat. This keeps each investigation focused and makes it easier to return to later.

Return to Previous Chats

Earlier conversations appear under Previous Chats in the AI navigation. Select a conversation to reopen its questions, charts, Insight, Analysis Summary and recommendations. The question box at the end allows you to continue from that context.

Use previous conversations for ongoing investigations—for example, revisiting a Channel after another week of data—rather than repeatedly recreating the same analysis. Start a fresh conversation when the project, definition or core question has changed materially.

Know when to use a report instead

Chat is strongest for exploration, comparisons and explaining a pattern. Use the underlying report or Explorer when you need a repeatable dashboard, exact filter control, routine monitoring or a figure that will be distributed as an agreed business metric.

Avoid including passwords, payment information or unnecessary personal data in a question. For lead-level investigation, use the User Dashboard under the organisation's normal data-access rules.