> For the complete documentation index, see [llms.txt](https://dots.gitbook.io/dots-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://dots.gitbook.io/dots-docs/reporting/reports.md).
# Reports
Reports **bring together multiple pieces of information** — qualitative and quantitative — into a single, coherent analytical view. Instead of looking at responses, annotations, and structured fields in isolation, a report lets you place them side by side, compare groups, surface relationships, and tell a clear story with your data.
A report is built on a blank canvas from small, reusable building blocks called widgets. Each widget answers one question about your data — a headline number, a distribution, a cross-tabulation, an AI-written summary, or a set of supporting excerpts. Arranged together, they turn a dataset into an argument your stakeholders can follow. Reports are especially useful for reporting, pattern analysis, storytelling, and sharing insight with stakeholders.
#### Why use Reports?
* **Combine qualitative and quantitative data in one place.** Structured survey fields, AI and NLP outputs, and raw transcript excerpts can all live on the same canvas.
* **Compare and cross-tabulate.** Break any measure down by group — caste, education, livelihood, sentiment — to see how findings differ across your respondents.
* **Let AI do the heavy lifting on qualitative data.** AI widgets summarise, categorise, and answer free-form questions across hundreds of transcripts in seconds.
* **Tell a story, not just show numbers.** Order widgets into a narrative arc, add framing text, and pair every claim with the evidence behind it.
* **Stay live.** Most widgets recompute automatically as new data comes in, so a published report keeps reflecting the latest dataset.
* **Share easily.** Publish a report to the collection page so stakeholders can view the finished analysis without touching the raw data.
### How to build your own report?
{% stepper %}
{% step %}
To start off a report, go to the ‘Reports’ menu from the left panel and click on ‘Create a New Report’.
{% endstep %}
{% step %}
Add a title and subtitle to to your report & 'Create Report'
{% endstep %}
{% step %}
#### Add new widgets for visualization
You will be directed to a blank canvas where you build your report using smaller components called widgets. Click the ‘+ Add Widget’ button on the canvas to open a dropdown of visualization options to choose from. Widgets are arranged on a 12-column grid, with up to three widgets per row, so you can lay related visuals side by side.
{% content-ref url="/pages/6ypUE3ViZ5XUoKduJNL3" %}
[More about Widgets](/dots-docs/reporting/more-about-widgets.md)
{% endcontent-ref %}
{% endstep %}
{% step %}
#### Publish the report
Once your report is finished, click the “Publish” button to save your changes. After publishing, you can return to the collection page, where all published reports are displayed and can be shared with stakeholders.
{% endstep %}
{% endstepper %}
#### Global filters
To apply a quick global filter all of the visualization on the canvas, use this filters tab. These filters allow you to customize the visual display of the data you wish to see.\
### Tips for getting the most out of Reports
1\. **Anchor the report to one research question.** A report built as a generic “dataset overview” tends to wander. Pick the single sharpest question you want to answer and structure the whole narrative around it.
2\. **Know your data before you design.** Check your field and tag frequencies first. Structured fields with real spread work well in charts and heat maps; broad themes that appear on almost every record will look nearly identical across groups and won’t differentiate anyone.
3\. **Match the widget to what actually varies.** When the difference lives in the numbers, use charts. When it lives in what people said rather than whether they said it, use AI Summary, Ask AI, or Annotation Explorer.
4\. **Use filters as your main lever** — especially on AI widgets. Two AI Summaries with the same prompt but different group filters, placed side by side, often reveal more than one broad open-ended question.
5\. **Pair every claim with its evidence**. Follow an AI-written insight with an Annotation Explorer or Responses List filtered to the relevant tag, so readers can see the actual excerpts.
6\. **Keep AI questions specific and groundable.** Vague questions get vague, hedging answers. Ask for examples, quotes, or specific groupings, and narrow the filter if the AI starts to speculate.
7\. **Use a Metric Card for “average of X by group.”** Bar charts count records — they don’t aggregate a separate numeric field. For “average income by education level,” build one filtered Metric Card per group.
8\. **Treat Visualize Table as a last resort.** It is static, so use it only for a finalised dataset or a one-off illustrative number — never while data is still being collected.
### Quick decision guide
* Need a single number? → Metric Card (filtered if it should be segmented).
* Need a distribution across one category with real spread? → Bar Chart or Pie Chart.
* Need to cross-tabulate two fields (or a field × a specific tag)? → Heat Map Grid.
* Need to compare what groups say, not just whether they said it? → AI Summary (paired and filtered) or Ask AI.
* Need to show the evidence behind a claim? → Annotation Explorer or Responses List.
* Have a date field and a real time dimension? → Line Chart.
* Have multi-step process data? → Sankey Chart.
* Already computed the numbers elsewhere and just need to show them once? → Visualize Table (remember it is static).