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

# Specific Topic Deep-Dive

> Use Smart Columns to Precisely Classify or Extract Topic-Specific Feedback

When analyzing open-text feedback, you may want to isolate comments related to a specific topic, such as pricing, onboarding, churn risk, usability, or billing.

While automatic topic assignment helps explore themes, some use cases require more control over how a topic is defined or extracted.

An LLM-powered Smart Column allows you to apply custom logic to either:

* **Classify comments** (e.g., Yes/No, labels)
* **Extract specific comments** related to a topic
* **Return structured outputs** based on defined criteria
  <iframe src="https://www.loom.com/embed/64509a3408874f3ab20ae427e041f108" title="Loom video player" frameborder="0" className="w-full aspect-video rounded-xl" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" allowfullscreen />

### Practical Use Cases

You might use a Smart Column when you want to:

* Detect **churn risk** only when cancellation intent is clearly expressed
* Flag **pricing complaints** only when dissatisfaction is mentioned
* Extract comments describing **implementation challenges**
* Identify feedback about a **specific process stage** (e.g., onboarding phase)
* Separate **product issues** from **service-related issues**

These scenarios often require stricter logic than topic clustering provides.

### Best Practices

<Steps>
  <Step title="Step 1: Define the Topic Clearly">
    Avoid vague instructions such as:

    > “Extract comments about pricing.”

    Instead, define the scope:

    > Identify comments that refer to pricing structure, subscription costs, perceived value for money, or price transparency.
  </Step>

  <Step title="Step 2: Add Inclusion and Exclusion Criteria">
    Example structured prompt:

    > Determine whether the comment refers to \[TOPIC].
    >
    >     <br />
    >
    >     <br />
    >
    > Include comments that clearly describe or evaluate defined elements of \[TOPIC].
    >
    >     <br />
    >
    >     <br />
    >
    > Exclude comments that mention the topic only incidentally or focus primarily on other themes.

    You can then configure the output as:

    * Yes/No classification
    * Summaries
    * Extracted comments
    * Structured categories
  </Step>
</Steps>