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

# Open-ended question results

> Working with the verbatims: word cloud, search, summary and coding.

An open-ended question is not read like a closed question: the answers are free text, the **verbatims**. The platform lets you explore them, summarize them and **code** them to make them usable.

## Exploring the verbatims

Two ways in to browse the answers without coding anything:

* **Word cloud**: click a word to display all the verbatims that mention it.
* **Search bar**: type a term to select all the verbatims that contain it.

## Generating an automatic summary

The **lightbulb** icon generates an AI summary of the question. It draws on the background and objectives entered when the questionnaire was programmed.

<Tip>
  The more precise the survey's background and objectives, the more relevant the summary. It is the same principle as a good prompt.
</Tip>

## Coding an open-ended question

**Coding** turns the verbatims into codes: categories you can count and cross.

### From a code frame

<Steps>
  <Step title="Open Coding">
    Go to the **Coding** tab of the question.
  </Step>

  <Step title="Define the codes">
    Three ways to define the codes: start from the items of another question (e.g. an awareness list), enter a list of labels, or reuse an existing code frame. Give your code frame a name, then confirm.

    For an open-ended question about **brands**, preferably start from the items of another question (your awareness list): the labels are already there.
  </Step>

  <Step title="Start the processing">
    The platform matches the verbatims to the codes using **spellings**: it automatically recognizes the spelling variants of one and the same label. The matching is never total — there is almost always a share of **not coded** verbatims left to finish by hand.
  </Step>

  <Step title="Handle the not coded verbatims">
    Go back to the verbatims left without a code. You can **reclassify** them into an existing code, create a **specific code** if one stands out (e.g. a competitor to watch), select the "I don't know" answers in bulk to make them a **DK** item, or group the rest into an **Other** item.

    You can also leave verbatims not coded: it has no real impact on the results, even if putting them in **Other** is cleaner.
  </Step>
</Steps>

<Note>
  The automatic matching covers a good share of the answers, not all of them. The remaining answers will then have to be handled manually.
</Note>

<Tip>
  Applying the **same code frame** to several questions lets you handle the not coded verbatims in one go.
</Tip>

### With AI

For long answers, the AI proposes a first sort by theme.

<Steps>
  <Step title="Choose AI coding">
    In the Coding tab, select **AI coding**.
  </Step>

  <Step title="Choose the level of codes">
    Indicate whether the AI should produce **One single level of codes** or **Two levels of codes (codes and subcodes)**. The two-level mode first brings out broad themes, then more detailed subthemes within each of them.
  </Step>

  <Step title="Review the proposed frame">
    The platform reads the verbatims and proposes a code frame, taking the survey objectives into account.
  </Step>

  <Step title="Adjust">
    Add a code that is useful to the client, delete the ones that are not needed, then confirm. You can also **Add a subcode** inside an existing code.
  </Step>

  <Step title="Let it run">
    The platform links the verbatims to the themes in a few minutes.
  </Step>
</Steps>

<Note>
  In **two-level** mode, the AI proposes codes and subcodes. In **one single level** mode, it sticks to the broad themes.
</Note>

### Reading a two-level coding in the results

When a question is coded on **two levels**, the results follow the hierarchy: each **subcode** appears after its **parent code**, with explicit numbering (1, then 1.A, 1.B…). The results of a first-level code **include those of its subcodes**.

## Combining several open-ended questions

To obtain **overall awareness** (a brand mentioned in at least one of two questions, for example top of mind + spontaneous), code both questions with the **same code frame**, then reimport an open-ended question that groups their verbatims. Since the verbatims are already coded, the platform finds the matches without recoding. See **Importing a variable**.

## Exporting the verbatims

You can export the list of verbatims in **Excel** format: one row per answer, with its interview ID and the associated verbatim.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.