> ## 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.

# Reviews space

> Explore your reviews in a 3D universe where closeness means similarity: comments that talk about the same thing group into topics

The **reviews space** is a 3D universe where each point
is a review with a comment. The position of each point is computed by
**semantic similarity**: reviews that talk about the same thing stay close
to each other and, as they accumulate, form **topics** you can explore. You
find it in the **Analysis** group of the panel, next to Statistics,
Data Studio and Reports.

Unlike a traditional chart, the space does not start from predefined
categories: topics emerge from what your customers write. It helps you
answer questions like "what are my customers talking about?" or "is this
problem new or has it been growing?".

## When it populates

The space builds itself, with no configuration:

* The **3D projection** appears as soon as a few reviews with a
  comment exist (just a handful is enough).
* **Topics** require volume: around **300 reviews with a
  comment** that are eligible. With less volume you will see the points, but without
  topic grouping.
* The computation is **updated automatically every night**, incorporating the
  new reviews.

Meanwhile, the view communicates the real state:

* **"We are preparing your space"**: the first projection is in progress
  and the view updates on its own when it finishes.
* **"There are not enough reviews yet"**: volume is missing to project. The
  view will populate as reviews with a comment arrive.

## Exploring the space

* **Zoom in and out**: when you bring the camera closer you see individual points; when you
  move it away, the points aggregate into stars with **exact counts**
  computed on the server, so the visual summary never lies.
* **Hovering over a point**: an excerpt of the comment of that review loads
  on the spot.
* **Clicking on a topic**: its card opens with the topic metrics.

## The topic card

When you select a topic, a card opens with **exact numbers, served
by the server topic catalog**, never estimated from the
3D drawing:

* Number of **responses** and **subjects** analyzed, with the breakdown
  of identified and anonymous customers.
* **First and last time** the topic appeared.
* **Trend**: rising, falling, stable or new.
* **Classification coverage**: what proportion of the topic reviews
  was classified.
* Score distribution across **critical, neutral and positive**.
* **Keywords** of the topic and **evidence reviews** that illustrate it.

## Filters, colors and table view

You can narrow down what is shown with four filters:

| Filter | Options                          |
| ------ | -------------------------------- |
| Period | Day, week, month, year, all-time |
| Source | woku, NPS, CSAT, CES             |
| Type   | Recognition, Improvement         |
| Score  | Critical, Neutral, Positive      |

The color of the points supports two modes: by **Topics** (each topic with its
color) or by **Feedback type** (recognition and improvement).

Besides the 3D space there is the **Table** view, which shows the same
information in list format. You can switch between the two whenever
you want, and if your browser does not support WebGL the table is enabled
automatically as a fallback.

## Approximate distances and reviews without a topic

<Info>
  The projection reduces a space of many dimensions to just three, so
  the **distances in the drawing are approximate**: they help you get oriented,
  not to measure. The numbers in the topic card are exact. The legend
  reminds you of this and also shows the real coverage of the view, for example
  "120 of 150 reviews projected".
</Info>

Reviews that do not match any topic are grouped as **"No
topic"**. It is normal for them to exist: they are atypical comments or very different
from the rest, and the space shows them instead of hiding them.