> ## Documentation Index
> Fetch the complete documentation index at: https://conveo.ai/docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Talk to Your Data

> Ask research questions, choose study scope, inspect supporting evidence, and create outputs with Conveo's assistant.

Talk to Your Data is Conveo's research assistant. Use it to explore a study, compare studies, ask follow-up questions, and draft research outputs. Answers can include text, charts, tables, and references to participant evidence.

<Info>
  Review [usage and limits](/docs/settings-and-management/talk-to-your-data-usage) for monthly allowances, billing, and overage controls.
</Info>

<a id="key-concepts-explained" />

<a id="pro-tips" />

## Quick Reference

| Job                          | Starting point                                                                                                              |
| ---------------------------- | --------------------------------------------------------------------------------------------------------------------------- |
| Explore one study            | Open the study's **Analyze** area.                                                                                          |
| Compare studies              | Start Talk to Your Data and select the relevant studies.                                                                    |
| Explore a research program   | Start a conversation from its [project](/docs/settings-and-management/projects-and-tags) or [Storyline](/docs/storylines/storylines). |
| Find relevant research first | Ask the assistant to find relevant studies; starting without an attached scope is available where enabled.                  |
| Supply a brief or reference  | [Attach documents or images](/docs/interview-data/assistant-context).                                                            |

<a id="complete-written-guide" />

## Step 1: Navigate to Talk to Your Data

Open Talk to Your Data from Conveo's main navigation, or start from a study's **Analyze** area. A conversation keeps its own history and associated studies. Reopen an existing conversation to continue its discussion, or start a new one for a different research question.

Viewing a conversation requires access to its associated studies as well as the conversation. Continuing or changing a shared conversation requires conversation editing access. See [sharing conversations](/docs/settings-and-management/permissions-and-sharing#sharing-conversations).

## Step 2: Select Your Studies

Select studies relevant to the question. For a new conversation started from the main Talk to Your Data entry point, Conveo can automatically select relevant studies from your first typed question. Review the attached studies before relying on the answer. Project and Storyline conversations start with their own scope.

Where a conversation starts with no studies attached, the assistant can help discover research and propose studies to add; it cannot analyze interview data until a study scope is attached.

In an existing editable conversation, open its context controls to add accessible studies or remove studies from the scope. The assistant can also propose additions for you to accept. Once studies are attached, the removal controls keep at least one study.

Changing scope updates the data available for subsequent analysis. It does not erase previous messages or guarantee that an earlier answer used the new scope. Start a new conversation if you need a clean separation from previous research context.

[Foundation studies](/docs/interview-data/foundations) shared by another organization are marked with their origin where available. They can support authorized read analysis, while some tools that save study changes are unavailable in conversations containing another organization's studies.

## Step 3: Craft Your Question

Name the decision, population, and comparison you need. Specify a format where it helps:

> Compare the main reasons for choosing our product among first-time and repeat buyers. Show themes, counts, and supporting quotes. Explain missing data and differences in the study questions before comparing results.

Follow up to inspect an outlier, narrow a group, or request the evidence behind a conclusion. Ask the assistant to state the denominator and whether counts refer to participants, interviews, diary entries, or mentions.

For background materials, use [assistant context](/docs/interview-data/assistant-context). Your [personal memory](/docs/settings-and-management/memory) can help tailor explanations. Neither background documents nor personal preferences replace participant evidence.

## Step 4: Launch and Wait

Send the question and follow the progress shown in the conversation. Data preparation and multi-step analysis can take time; there is no fixed completion time. The assistant retrieves and queries relevant data, rather than guaranteeing that every question reads every transcript in full.

Use the stop control if you want to stop the current response. If a turn fails, inspect the displayed error and the conversation before retrying. A partially completed answer may need follow-up.

Long conversations may summarize older messages to remain within the context limit. If prompted, use **Start next chat from summary** to continue in a new thread. Review the transferred context when precise wording or an earlier constraint matters.

### Retry the latest question

To run your latest question again, hover over your message and select **Retry message**. A response that was cut short can also offer a retry control. Retrying regenerates the latest answer without adding another copy of your question; the previous answer is replaced in the displayed conversation.

Retry requires conversation editing access and is unavailable while an answer is running or an account limit blocks the conversation. Earlier questions do not have this control. After manually editing or renaming an output, retry may be unavailable until you send another question. To preserve an alternative direction, use [conversation branching](#explore-a-different-direction).

## Step 5: Review Your Results

Open evidence references and compare claims with their source responses. Check cohort filters, question wording, missing values, and sample sizes before sharing a finding. Generated charts and counts reflect available data and the selected analysis, not automatically a representative population estimate.

The assistant can use web research and attached background materials as well as study data. Distinguish external context from findings supported by Conveo interviews, and ask for source clarification when needed.

Use dedicated output workflows for durable deliverables:

* [Reports](/docs/insights-and-reports/reports) for structured written findings.
* [Slide decks](/docs/insights-and-reports/slides) for presentations.
* [Quote reels](/docs/insights-and-reports/quote-reels) for participant video evidence.

Review and save or publish the intended output through its controls. A useful answer in a conversation is not automatically a published report.

<a id="best-practices" />

## Explore a different direction

Use **Branch conversation from here** beneath an assistant response to start a new conversation containing the history up to that response. Later messages remain in the original conversation. The branch starts with the source conversation's studies and project context; review its scope and sharing before continuing.

Branching requires access to the source conversation and its studies. It does not edit the original conversation or make its evidence public. Use the new conversation for a different question or interpretation while keeping the original discussion intact.

## Edit a generated document

When a response opens an editable document, use its editing and title controls to revise the content. Editing requires conversation editing access. Check the save state and finish with **Done** before leaving; if saving fails, retry before navigating away.

A manual edit changes the conversation's document, not the underlying interview evidence. Review any published or exported copy separately rather than assuming a chat edit updates every deliverable.

## Limitations and troubleshooting

| Situation                             | What to do                                                                                                      |
| ------------------------------------- | --------------------------------------------------------------------------------------------------------------- |
| A study is absent                     | Check your organization and [study access](/docs/settings-and-management/permissions-and-sharing).                   |
| An answer lacks support               | Ask for source quotes or a data table; inspect the underlying interviews.                                       |
| Studies have incompatible questions   | Compare the concepts explicitly and report the limitation instead of assuming equivalent measurements.          |
| Data is still preparing               | Wait for the displayed preparation state, then continue.                                                        |
| A question is blocked by usage limits | Review [usage and limits](/docs/settings-and-management/talk-to-your-data-usage).                                    |
| A document or image is missing        | Review [attachment requirements](/docs/interview-data/assistant-context) and wait for parsing or upload to complete. |

AI analysis can omit evidence, misinterpret wording, or overstate a pattern. Verify consequential findings before making decisions or sharing them.