> ## Documentation Index
> Fetch the complete documentation index at: https://docs.coloop.ai/llms.txt
> Use this file to discover all available pages before exploring further.
# Analysis Grids
> Build an analysis grid: choose which participants or segments each row covers, ask questions as columns, read the overview across rows, and trace every answer back to the transcript.
An analysis grid turns your research material into a table you control. You decide what each row spans, what each column asks, and the type of output you'd like, and CoLoop generates the answers, with citations back to the source.
* **Rows** decide whose data an answer draws on — one participant, one segment, or a combination you build.
* **Columns** are the questions you ask. Each column has an output type, so every answer fits the question: synthesized analysis, verbatim quotes, an extracted number, a yes/no, or coded categories with counts.
* **Cells** hold one row's answer to one column's question.
* The **Overview** row at the top gives a summary of all rows' responses to each question.
## Create a grid
1. In your project sidebar, click **Analysis grid**.
2. Click the **+ Analysis grid** card.
3. Give the grid a name.
4. Choose how to set it up:
* **Custom grid** — build the grid from your project data. Pick where the rows come from (participants, segments, or an existing grid), then where the columns come from (your discussion guide, an existing grid, or skip to add your own questions).
* **Start from scratch** — create an empty grid and add the rows and columns yourself.
5. Click **Create grid**.
With a custom grid, CoLoop creates the grid and then prompts you to select the exact participants, segments, and questions you'd like in your grid. On your very first analysis grid, the [display setup](#change-how-the-grid-displays) comes first, and then you will be prompted to select your row and column setup once you've saved or skipped it.
A row or column source you can't use yet is grayed out: **From participants** and **From segments** require participants or segments in the project, **From discussion guide** requires a discussion guide, and **From existing grid** requires another grid in the project to copy from.
## Choose your rows
Each row carries a filter that decides whose data its answers come from. A row can be a single participant, a whole segment, or several of both combined.
### Add one row per participant or segment
1. Click **Add Participant Rows** in the grid header.
2. Stay on the **Add many** tab.
3. Search for participants and segments, and select the ones you want. Use **Add all** on a group heading to select everything in it, and **Clear all** to undo that.
4. Click **Add rows**.
Everything you select becomes its own row. CoLoop skips selections that duplicate a row already in the grid or that contain no participants, and tells you how many it skipped.
### Combine several participants or segments into one row
1. Click **Add Participant Rows** and open the **Build custom row** tab.
2. Select a segment or participant, then click **+** to add another.
3. Click the **AND** or **OR** between the entries to change how they combine.
4. Click **Add row**.
Use **OR** to expand a row so it covers everyone in either group. Use **AND** to narrow it down to only the participants who belong to every segment listed. **AND** only works with segments — a row that names a specific participant has to use **OR**.
Turn on **Create more** to keep the builder open so you can add several custom rows without reopening the dialog.
### Manage rows
* **Add a row inline** — use the **Add row** cell at the bottom of the participant column, which has the same builder as the dialog.
* **Reorder** — drag a row by the handle on its left edge. Remove any sorting first; rows can't be dragged while sorting is applied.
* **Edit** — hover over a row, open its three-dot menu, and click **Edit row** to change which participants and segments it covers.
* **Delete** — use **Delete row** in the same menu, or select several rows with their checkboxes and use **Delete** in the header bar.
* **Show segments** — each row lists the segments its participants belong to. Move them into their own column from the **Display** menu.
### Rows that lose a participant or segment
A row remembers the exact participants and segments you gave it. If one of them is later deleted from the project, the row labels it **Removed participant** or **Removed segment** in place of the name.
Two things commonly cause this:
* **Retranscribing a file.** Retranscribing replaces that file's speakers with new ones, so any row built on the old speakers no longer finds them.
* **Deleting a segment** that a row was built from.
A row that still covers someone keeps working, and its answers come from the participants that remain. A row that has lost everyone can't produce answers at all, and recomputing the grid won't bring them back — the data it pointed at is gone. To clear it:
1. Hover over the row and open its three-dot menu.
2. Click **Delete row**.
3. Add a new row for the participants or segments you want to analyze instead.
You can also use **Edit row** to drop the removed names and keep the rest of the row.
Grid exports carry the same labels, so a row that lost a participant reads the same way in your spreadsheet as it does on screen.
## Ask questions as columns
Each column is one question, asked of every row.
1. Click **Add a Question Column** at the bottom of the grid, or press **⌘K** (**Ctrl+K** on Windows).
2. Type your question.
3. Choose the output type, and add labels if you picked **Categorization**.
4. Press **Enter**, or click **Add column**.
CoLoop names the column from your question and starts computing straight away. The question composer stays open so you can add the next question immediately. Each column you add counts as one question against your plan's usage.
The column name is a short stand-in for your question. To read each question in full under its column title, turn on **Column questions** in the [display settings](#change-how-the-grid-displays).
### Output types
| Output type | What each cell returns | Use it for |
| - | - | - |
| **Synthesis** | A written answer with citations | Themes, summaries, reasoning, anything open-ended |
| **Quotes only** | Verbatim quotes from the participant | Pull-quote banks and quote sheets |
| **Numeric** | A single number, with an explanation | Ratings, prices, counts, ages |
| **Yes / No** | Yes, No, or Undetermined, with an explanation | Screening a behavior or an attitude across everyone |
| **Categorization** | Two or more labels you define | Coding responses against a codebook, participant counts per category |
A **Categorization** column needs at least two labels — the set CoLoop chooses from when it codes each row. Pick **Categorization** in the composer and the **Labels** row appears, where you can either:
* **Ask CoLoop for a starting set** — click **Suggest labels** and CoLoop drafts a set from your question. Write the question first: the button stays disabled until there's enough to work from.
* **Write your own** — click **Add your own**, then the **+** to add each label after that.
In both circumstances you will end up with a chip per label. Click a chip to rename it, remove it, or give it criteria describing when the label should apply — criteria are optional, but they sharpen the coding. Select **Mutually exclusive** when each row should get exactly one label.
Label names can't contain quotation marks, backslashes, or line breaks.
**Suggest labels** disappears once a label exists, so to re-draft a set, remove the labels you have. For a bigger codebook, open **Advanced**: the full editor lists each label with its criteria on its own line, and accepts a comma-separated or line-separated paste to add many at once — any pasted label with an unsupported character is skipped, and CoLoop tells you how many were left out.
### Narrow a column to specific data
Click **Add filter** in the composer menu to limit the question to certain files and concepts. Leave it empty to analyze all of the project data. This is per column, so one grid can hold a question about a single concept next to a question about several concepts.
Row filters and column filters do different jobs: row filters decide *whose* data is analyzed, and column filters decide *which of their data* is analyzed.
### Set a custom overview, or skip it
Click **Advanced** in the composer menu, or **Show advanced** in the add-column dialog, for the per-column overview controls:
* **Overview prompt** — replaces your question for the Overview row only. Use it to say what the summary across rows should answer and how it should be formatted. Leave it empty for the automatic thematic overview. Available on **Synthesis** columns only.
* **Skip overview** — leave the Overview cell empty for this column. Available on every output type, and off by default, so every column gets an Overview. On a **Quotes only** column the Overview is a synthesized thematic summary of the quotes across rows, while the per-row cells stay verbatim.
### Add columns from your discussion guide
Suggested questions come from the discussion guide you added during project setup. CoLoop intelligently groups these questions based on similar themes present in the guide.
Open the composer menu and, before you type anything, CoLoop lists suggested questions above the input. Click one to add it as a column, tick several and add them together, or click the pencil to load a question into the composer and change it first.
For the full selection of questions, click **Advanced**, then **Create from discussion guide**:
1. Choose questions from the **Suggested** tab, or from **Extracted** for the exact questions written in your discussion guide.
2. Click a question to review it on the right, and adjust its title, prompt, output type, labels, or filter before adding.
3. Every selected question will be added as a column.
To give a whole batch the same output type, set it on one question then click **Apply … to all selected**.
Questions that are already asked in columns are not suggested again. A question that can't be added as it stands — a categorization question with no labels, for example — is flagged so you can fix it first.
### Manage columns
Click a column header to open its menu:
* **Edit column** — change the question, output type, labels, filter, or overview settings. Editing recomputes the column.
* **Insert left** / **Insert right** — add a new column next to this one.
* **Duplicate column** — copy the question and its settings into a new column to tweak.
* **Delete column**.
Drag a column header to move the column, and drag its right edge to resize it.
## Read the results
Cells compute as soon as you add or edit a row or column, and recompute whenever you change something that affects them. You don't need to trigger anything.
While a grid is working, cells fill in as their answers land. Cells load as you scroll, so a large grid still opens quickly.
| Cell shows | What it means |
| - | - |
| A loading placeholder | The answer is being computed |
| **Failed** with **Retry** | The cell couldn't be computed. **Retry** re-runs every failed cell in the grid |
| *No data to analyze. The current filters exclude all matching responses.* | Nothing was in scope for this cell. The column's file and concept filter and the row's participant or segment selection don't overlap, so there was nothing to read. Expand the column filter, or change what the row covers |
| **Undetermined** | A Yes/No column where the data doesn't answer the question either way |
| **No label** / **No labels** | A categorization column where none of your labels apply. A multi-label column shows this instead of the no-data message when a filter excluded everything, so check the column's filter before concluding your labels didn't fit |
| **Skipped** | An Overview cell for a column which skips its overview |
| **Not computed** | An Overview whose participants changed outside the grid, so no answer was ever built for the current set. Hover the cell and click **Compute grid** |
### The Overview row
The Overview row answers each column across every row in the grid. It appears once the grid has more than one row. When no row in a column had anything in scope, the Overview carries the same no-data message its cells do.
What you see depends on the output type:
* **Synthesis** columns get a thematic breakdown with the number of participants behind each theme — or your own summary if you set an overview prompt.
* **Numeric** columns get a single reduced value, such as an average or a total, with an explanation backing this value.
* **Yes / No** and **Categorization** columns get a distribution across participant of the count and percentage for each value.
#### When an Overview reads "Not computed"
An Overview answer covers an exact set of participants. Changes that happen **outside the grid** leave an Overview with no answer. Adding or removing a participant from a segment a row is built on, or a participant disappearing when their file is retranscribed, both move who the Overview covers. It then reads **Not computed**, rather than showing an answer that no longer matches the rows beneath it.
Hover the cell and click **Compute grid** to produce it.
That button computes the **whole grid**, not just the one Overview. It is the same action as **Compute grid** in the grid header, so it also picks up anything else outstanding — cells that never computed, or that failed.
### Look at one cell or one row in detail
Click any cell to open the details panel. It shows the row's filter, the full untruncated output, and — for Yes/No and categorization cells — the participant distribution. Click a bar in this distribution to see the individual participants' answers determining that value.
After selecting a cell, use the arrow keys to move to the next cell in that direction. The details panel follows your selection. If focus is on a control inside the cell, such as a citation, the arrow keys continue to control that item.
To read one participant's responses across every question, hover over their row and click **Open**. The row panel lists every column's answer for that row.
### Check the evidence
Every generated answer is cited. Click a citation to open the **Citation sources** panel, expand a quote to read it in context, click **Open** to jump into the transcript, or turn the quote into a clip.
For more on working with evidence and clips, see [Evidence Panels](/docs/analysis/evidence-panels) and [Creating Clips and Reels](/docs/analysis/creating-clips).
## Sort and filter what you see
Hover over a column header to reach its sort and filter icons, or add a rule from the bar above the grid.
* **Sort** — sort rows by any column, ascending or descending.
* **Filter** — filter by participant, by segment, or by the values in a **Yes / No** or **Categorization** column. Yes/No filters by Yes, No, or Undetermined; Categorization filters by label.
* **+ Filter** adds another rule, and **Reset** clears every rule.
Filtering needs a known set of values to pick from, so it's offered on **Yes / No** and **Categorization** columns only. **Synthesis**, **Quotes only**, and **Numeric** columns are sort-only: they have no filter icon and don't appear in the **+ Filter** menu. If you edit a column from Yes/No or Categorization into one of those types, any filter on it is dropped.
Sorting and filtering only change your own view of the grid. Nothing is saved to the grid or shown to collaborators.
### Overviews for a filtered view
A column overview summarizes every row, so it would misrepresent any filters applied. When a filter narrows the rows, the Overview row becomes **Filtered overview** and its cells read **Not computed** until you ask for them.
Click **Compute** on the Filtered overview row, or on an individual cell, to summarize exactly the rows you're looking at. Results are kept, so returning to the same filter shows the summary again without recomputing.
Some subsets can't be summarized, and the cell says why:
| Cell shows | What it means |
| - | - |
| **Too many rows to summarize** | Narrow the filter further |
| **A row failed - please retry** | A row's cell in that column failed. Retry it, then compute again |
| **These rows can't be combined** | The filtered set matches an existing row too closely to summarize separately |
## Recompute a grid
Grids recompute themselves as you edit them, so you rarely need to do this by hand. When you do — after a run of failures, for example — open the three-dot menu in the grid header and click **Compute grid**. CoLoop picks up anything that hasn't succeeded and leaves finished cells alone. If there's nothing outstanding, it tells you so.
**Retry** on any failed cell re-runs every failed cell in the grid, so you don't have to hunt them down one at a time.
## Change how the grid displays
Display preferences are saved to your account, so they will be set as the default on any device you use and apply to all grids in every project.
The first time you open an analysis grid, CoLoop allows you to choose these default display settings. **Welcome to Analysis Grids** presents each preference next to a mock grid that shows the change as you make it. Click **Save** to keep your choices, or **Skip for now** to start with CoLoop's default settings.
To change them later, open **Display** in the grid header. Editing a setting there applies immediately. For the descriptions and the mock grid, click **Display settings** at the top of that menu — choices in the panel are written when you click **Save**, and **Cancel** reverts whatever you changed.
| Preference | What it does |
| - | - |
| **Expand cells** | Show each answer in full instead of limiting it to the first few lines |
| **Show scrollbars** | Keep the grid scrollbars on screen rather than fading them in while scrolling |
| **Column questions** | Show each column's full question underneath its title |
| **Column previews** | While you're adding a question, outline where its column will land |
| **Quote context** | Show the researcher turn or task prompt above each quote in Quotes only columns |
| **Segments: Show as column** | Break segments out of the participant cell into their own pinned column |
| **Segments: Expand all** | Show every segment a row belongs to, instead of collapsing them |
| **Evidence panel: Expand citations** | Open each citation with its quoted text showing, rather than collapsed to the source |
### Fit more on screen
When a grid has many columns, three controls give you more room. These stay on the device you set them on rather than following your account.
* **Zoom**: scale the whole grid from 20% to 160% with the minus and plus buttons in the **Display** menu.
* **Full screen**: hide CoLoop's navigation so the grid fills the window. Click the header button again, or leave the grid, to bring it back.
* **Collapse the participant columns**: click the collapse arrow on the **Participants** header to fold the left columns into a thin rail. Click it again to expand.
## Export, rename, and share
Open the three-dot menu in the grid header for:
* **Export** — download the grid as an Excel file (XLSX), with a column of participants, their segments, the Overview row, and one column per question.
* **Manage access** — control who can see and edit the grid. See [Sharing analysis grids and AI chats](/docs/collab-and-access-management/sharing-analysis-grids-ai-chats).
* **Rename**.
* **Delete**.
You can also copy any cell or column out of the grid and paste it into Excel, Word, or a deck.
## Work across several grids
You can create as many grids as you like in one project — one per topic, workstream, or market for example. Click the grid name in the header to switch between them, and use **From existing grid** when creating a grid to reuse another grid's rows or columns as a starting point. A categorization column whose labels contain unsupported characters can't be copied — CoLoop leaves it out and counts it among the skipped columns.
## Coming from an older grid
Basic and comparative grids are now read-only. Open one and click **Clone to new grid** to copy its rows and columns into a new analysis grid, which then computes fresh with this updated format.
For the older grids themselves, see [Basic Grids](/docs/analysis/basic-grid) and [Comparative Grids](/docs/analysis/comparative-grids).
## Writing good questions
The question you write drives everything in the column. For prompting advice that applies across grids and chat, see [Top tips and prompting advice](/docs/analysis/top-tips-and-prompting-advice).
This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.