> For the complete documentation index, see [llms.txt](https://docs.askyazi.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.askyazi.com/results-and-reporting/analysis-limits.md).

# Analysis limits

What statistical analysis Yazi does and does not do today, and how to handle weighting, significance, banners and sample sizes.

Yazi covers the everyday analysis most studies need: charts, filters, cross-tabs, themes and reports. Some quantitative work still happens in your own tools. This page says plainly what is and is not built, and what to do instead.

{% hint style="info" %}
**In short:**

* **Weighting is not built in.** Export the data and weight it in SPSS, Q or similar.
* **Cross-tabs split by one question at a time.** There are no multi-column banners.
* For significance testing, use the [AI assistant connector](https://docs.askyazi.com/integrations-and-api/connect-claude-or-chatgpt.md) (chi-square on cross-tabs) or your own tools.
  {% endhint %}

## Quick reference

| Need                                                 | In Yazi today                    | What to do                                                                                                                                                                                                   |
| ---------------------------------------------------- | -------------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| Weighting                                            | Not available                    | Export (CSV with numeric codes, or SPSS) and weight in SPSS, Q, R or Excel. Use quotas during fieldwork so your sample needs less weighting.                                                                 |
| Significance testing                                 | Not shown on the Graph Data page | Ask through the AI assistant connector, which runs a chi-square test on a cross-tab, or test in your own tools.                                                                                              |
| Banner tables (several break variables side by side) | Not available                    | Use **Cross-tab > Just this session** to split every chart by one question, or export and tabulate elsewhere.                                                                                                |
| Ranking questions                                    | Not available as a question type | See [Question types](https://docs.askyazi.com/build-a-study/question-types.md) for the workaround using piping.                                                                                              |
| Charting AI interview follow-ups                     | Not available                    | Read them in **Transcripts** and see their themes in **Insights**.                                                                                                                                           |
| Live dashboard in Power BI or Tableau                | No built-in connector            | Use [Google Sheets or Excel](https://docs.askyazi.com/integrations-and-api/google-sheets-and-excel.md), [Response Push](https://docs.askyazi.com/integrations-and-api/response-push-webhooks.md) or the API. |

## Weighting

Yazi does not weight results. Every chart, cross-tab, report and export is unweighted. To weight:

1. Export from **Table Data** as CSV with **Numeric codes** and **Short variable names**, or as SPSS (.sav) for up to 10,000 rows.
2. Apply your weights in SPSS, Q, R or Excel.

Quotas set during fieldwork give you the right shape of sample to start from. See [Quotas](https://docs.askyazi.com/build-a-study/quotas.md).

## Significance testing

Charts and cross-tabs on **Graph Data** show counts and percentages, without significance markers. To test significance, export the data or use the AI assistant connector.

The AI assistant connector can cross-tabulate two closed questions and include a chi-square test of whether they are related. Cells with fewer than 5 people are hidden in its output so individuals cannot be identified, but they still count towards the test. See [Connect Claude or ChatGPT](https://docs.askyazi.com/integrations-and-api/connect-claude-or-chatgpt.md).

## Sample size guidance

Yazi follows the standard rules of margin of error and statistical significance. There is no Yazi-specific method. Some working numbers:

* A national read from one sample needs a few hundred completes. Around 300 to 500 is a common target for a representative read.
* If you want to compare segments, each segment needs its own base. Around 30 to 50 per segment is a practical minimum for directional reads, and more for confident comparisons.
* On a representative sample with quotas met, around 1,500 per cohort is plenty. Going far above that changes the precision very little.
* A national sample of a few hundred does not give a reliable read on each of your sub-groups. Plan sample by segment from the start if you need segment results.

For hard-to-reach audiences, describe the base honestly rather than presenting it as representative.

## Limits and things to know

* Everything on screen is unweighted.
* Cross-tabs use one break variable at a time.
* AI-written summaries, themes and reports are a first draft. Check them against the data.

## Related

* [Graph Data and cross-tabs](https://docs.askyazi.com/results-and-reporting/graph-data-and-cross-tabs.md)
* [Exports](https://docs.askyazi.com/results-and-reporting/exports.md)
* [Quotas](https://docs.askyazi.com/build-a-study/quotas.md)
* [What Yazi cannot do](https://docs.askyazi.com/help/what-yazi-cannot-do-yet.md)
