> For the complete documentation index, see [llms.txt](https://evidano.gitbook.io/ailyze-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://evidano.gitbook.io/ailyze-docs/analyze-data/design-your-analysis-table.md). # Design Your Analysis Table Shape the table a document analysis produces: choose what each row is, pick column types, add themes, questions, and segments, and set the report. This page shows researchers how to review and shape the table that a document analysis drafts for them before the analysis starts. By the end you will know what each row represents, which column type gives you the output you need, and how themes, questions, segments, and report options fit together. Every control is listed on [Document Analysis Settings](/ailyze-docs/reference/settings/document-analysis.md#screen-6-your-table-the-table-builder). The examples use the guide's sample project: three staff interviews and one parents' focus group evaluating a mobile health clinic. ## Before you start * Follow [Analyze Interview Transcripts](/ailyze-docs/common-research-workflows/analyze-interview-transcripts.md) up to the **Your table** step, so the AI has drafted a table. * Know the one or two things your report must answer. The table should serve them, not list everything the files contain. ## How the table works Every analysis is a table. Each **row** is one unit you want to learn about, such as one interview, one focus-group participant, or one outcome. Each **column** is one thing the AI extracts from every row: a theme, an answer, a category, a number. The report, charts, and comparisons are all built from this table. The draft shows it in three sections: **You will receive a report** (what the report includes), **You will receive themes generated from the entire dataset** (the theme columns), and **You will receive a table of answers and metadata** (every other column).
Table of answers and metadata with the unit sentence, Reorder, Add, two segment columns and a Text column above sample rows

The sample rows under the headers show the shape of the output, not real results.

## Step 1: Check what each row is Under **What is your unit of analysis?**, read the sentence **Each row is one …**. Click the word after it to change the unit. | Your material | A good row | Why | | ------------------------------------------- | ------------------------------ | ---------------------------------------------------------------------- | | One-to-one interviews | One document | Each interview is one account. | | Focus groups | One focus-group participant | Counts by person, not by transcript. Set on the **Focus groups** step. | | Evaluation reports listing several outcomes | One outcome (pick **Outcome**) | Each outcome is judged separately. | | Papers in a literature review | One document | One paper per row, for data extraction. | You should see the sample rows change to match the new unit.
What is each row popover with A document or focus-group participant, the example units and Something else

A table has one unit: a document, a focus-group participant, a mix of the two, or something the documents describe, such as an outcome.

## Step 2: Review the drafted columns In **You will receive a table of answers and metadata**, click each column (or its pencil, **Edit column**) and read its description: it is the instruction the AI follows for every row. Rename columns to the words you will use in your report, and remove any you will not use. **Undo** and **Redo** in the builder bar cover every change, so you can try a column and take it back. When you chose a method, the draft keeps that method's core columns and holds the rest under the table, in **More columns for** followed by the method's name. Open it and click a column to add it, or click **Add all**. Added columns go after the table's other columns. Literature-review methods (data extraction, quality appraisal, reporting checklists) keep their full set in the draft.
More columns for Framework analysis, open under the table, with one button per column and Add all

Hover a column's button to read what it extracts before you add it.

## Step 3: Choose the right column type The type decides what you can do with the answers later. Text is easy to read but cannot be counted; choice columns can be counted, charted, and compared. | You want to… | Use | Sample column | What one row might hold | | --------------------------------------------- | ----------------------------------------- | ----------------------------------------- | --------------------------------------------------------------------- | | Describe something in a sentence or two | **Text** | Account description | "Clinic manager responsible for two vans in South Ridge." | | List several short items | **List of text** | Suggested improvements | "Backup phone line; supply buffer for test kits" | | Put each row in one category | **Single choice (pick one)** | Overall tone | Mixed | | Tick every category that applies | **Multiple choice (pick all that apply)** | Barriers named | Fuel costs; Breakdowns | | Answer an open question and count the answers | **Question & Code** | What barriers does this account describe? | A short answer, coded as "Unreliable schedule" and "Supplies run out" | | Pull a stated figure | **Number** | Villages served | 11 | | Pull a date | **Date** | Interview date | 4 March 2026 | | Answer yes or no | **Yes / No** | Mentions privacy concerns | Yes | These sample values are illustrative, written for the guide's sample files. For choice columns, choose **Specify my own options manually now**, type one option per row, and click **Add descriptions** if an option needs explaining.
Single choice column window with five options and descriptions, Use both, Add option and the import buttons

Option descriptions are optional; the AI uses them to decide which option fits.

## Step 4: Set up the themes Themes, sub-themes, and codes are generated together, in **You will receive themes generated from the entire dataset**. The AI often drafts them as the builder opens. If the section shows **Configuration required**, or is missing (then click **Add** › **Add theme**), choose one of: 1. **Generate draft themes and sub-themes now**: the AI reads your files and proposes a set. Edit the pre-filled instructions to steer it, for example "Focus on barriers to using the clinic". 2. **Add manually**: **Type**, **Paste**, **Upload**, or **Use previous** to bring your own codebook. You should see a numbered list of themes and sub-themes. Click **Edit themes and sub-themes** to rename, merge, or delete them, then **Done**.
Themes section with numbered draft themes and sub-themes, Edit themes and sub-themes, and the sample table

Rename and merge draft themes now; it is quicker than editing the codebook after the run.

## Step 5: Add your research questions Each question becomes a **Question & Code** column. Click **Add** › **Add question**, paste your questions, click **Convert into list**, untick any you do not want, then **Add N questions**. Questions already in the table are left out. Phrase each one as you would ask it of a single document, for example "What barriers to using the clinic does this account describe?" rather than "Which district has more barriers?"; comparisons come from segments. ## Step 6: Add segments to compare groups A segment is a group you want to compare, such as district or stakeholder type. If the draft has segments, keep **Compare groups (e.g., by gender, year)** ticked. The AI tags every file as the draft opens ("Tagging files with AI…"); click **Review** in its message to check, or **Configuration required** to tag any it missed (**Auto-populate with AI**, or click a cell to add a value). **Add** › **Add group to compare** adds another. See [Compare Groups (Segments)](/ailyze-docs/analyze-data/compare-groups-segments.md).
Group tagging dialog listing each participant and document with a Respondent group cell

Amber warnings at the top list anything still untagged.

## Step 7: Choose what the report includes At the top of the draft, under **You will receive a report**, keep **Executive summary** and **Summary of each document** on, set **Report in** to your language, and turn on **Research questions mapped to themes** when each question should be answered through your themes. The outputs every project gets are folded under **Also included:**; click it to see them. You should see no amber **Configuration required** left. Click **Review and start analysis**; if anything is still missing, a message names it.
Optional: common mistakes and how to avoid them * **Too many columns.** Every column is extracted for every row; a focused table gives a sharper report. Add a method's extra columns from **More columns for** only when you need them. * **Vague descriptions.** "Barriers" alone invites guesses; "Obstacles to attending the clinic that the speaker describes, such as distance or timing" does not. * **Text where you need counts.** If you will want a chart, use a choice or question column. * **Comparing groups with a question.** Ask questions of each document; let segments do the comparing. * **Untagged files.** A file without a segment value drops out of that comparison.
## Next steps * Run it — [Analyze Interview Transcripts](/ailyze-docs/common-research-workflows/analyze-interview-transcripts.md) walks the whole project from upload to report. * Bring your own codebook — [Build a Codebook](/ailyze-docs/analyze-data/build-a-codebook.md) explains theme hierarchies in depth. * Every control — [Document Analysis Settings](/ailyze-docs/reference/settings/document-analysis.md) lists each option in the builder.