> ## Documentation Index > Fetch the complete documentation index at: https://woku.app/docs/llms.txt > Use this file to discover all available pages before exploring further. # Actionability and sentiment > The two readings woku makes of every comment: whether someone has to act, and how positive it is. Every comment received by a woku, NPS, CSAT or CES is analysed on two independent dimensions. Each one answers a different question and is used for something different. | | Actionability | Sentiment | | - | - | - | | Question | Does someone have to do something with this comment? | How positive is the comment? | | Values | Requires action (opportunity) or no action required (recognition) | Positive, neutral or negative | | Used for | Decides the follow-up agent, alerts, tickets and action plans | Analytics: statistics, reports and cards | | When it is computed | When the comment arrives, before any action | Afterwards, without blocking anything | ## Actionability: recognition and opportunity A comment **requires action** (opportunity) when it asks for, suggests or implies something the company should do, fix, change or answer. This includes complaints, suggestions inside praise ("great talk, but it ran long") and questions that expect an answer. A comment **requires no action** (recognition) when there is nothing to do: praise, thanks or reflections, even when they use words such as "failure" or "problem" as a topic. Actionability does not measure whether the comment was positive or negative. A five-star comment with a suggestion requires action, and that is correct. Comments with no content (rating only, empty text or a placeholder sent by an integration) are not classified and never trigger an action. ## Sentiment Sentiment describes the tone of the comment: positive, neutral or negative. It is analytics only and never triggers actions. ### Sentiment of a comment The AI does not give each comment a single label: it estimates what share of readers would read it as positive, neutral and negative (the three add up to 100%). This way a mixed or ambiguous comment is not forced to one side. * The comment's **label** is the reading with the highest share (on a tie, neutral). * The comment's **number** is the positive share minus the negative share, from -100 to +100. Example: a comment read as 20% positive, 30% neutral and 50% negative is labelled negative, with a number of **-30**. A calm complaint weighs less than an angry one, which could land at 0% positive, 5% neutral and 95% negative (**-95**). ### Net sentiment For a set of comments (a woku, a folder, a tool or a period), woku shows: * The **distribution**: positive, neutral and negative shares, adding up to 100%. It is the average of the shares of each comment. * The **net sentiment**: positive share minus negative share, with neutrals included in the total. It ranges from -100 to +100 and is shown without a percent sign, like NPS. Example: 60% positive, 25% neutral and 15% negative give a net sentiment of **+45**. Calculation rules: * Each comment counts once and contributes its three shares. The sentiment of a folder or a period is always computed over all of its comments, never by averaging its parts. * With fewer than 10 analysed comments, woku shows the distribution but not the number, because the sample is small. * Text sentiment is different from the star average. They are shown side by side. ## In the API `/v1` responses that include comments carry both fields: | Field | Values | Description | | - | - | - | | `feedbackType` | `improvement`, `recognition` or `null` | Actionability. `null` when there is no content to classify or the classification has not finished. | | `sentiment` | `positive`, `neutral`, `negative` or `null` | Text sentiment. `null` until analysed. | This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.