What is a Model?

3 mins read - Updated on Oct 28, 2026

A Model in Kimola is an AI model that reads every piece of customer feedback and adds structure to it: the themes the feedback mentions, how the customer feels, the answer it gives to a research question or the names it contains. Kimola has three kinds of models. Classifiers tag feedback with labels, Dimension models answer an analytical question about feedback, and extractors pick out entities. Every Report and every Space is analyzed with models, so the themes, sentiment and Dimension answers you read there come from them.

Every model is also either pretrained or custom. Pretrained models are ready-made by Kimola and work without any setup. Custom models are the ones you create from a prompt, a training dataset or a Report, with your own labels. The labels a model assigns are what make thousands of reviews, survey responses and support conversations countable, comparable and filterable.

Classifiers

A classifier tags every piece of feedback with labels. Its labels are the tags it can assign, and in a custom model each label has a name and a short description of what it covers. Kimola has two kinds of classifiers.

Content classifiers sort feedback into the topics that matter to your business, such as Delivery Time, Pricing or Mobile App. In Reports and Spaces, their labels appear as themes. The Automatic Classifier, a pretrained content classifier, finds the themes in the data itself and names them after what customers say, so it needs no list of labels. Industry classifiers such as the E-Commerce Feedback Classifier come with labels of their own, and a custom content classifier uses the labels you define.

Sentiment classifiers detect how customers feel in each piece of feedback. The pretrained Sentiment Classifier works on any kind of feedback. A custom sentiment model always uses two labels, Positive and Negative, and you can edit their descriptions to say what counts as each for your business.

Whether one piece of feedback gets one label or several is decided by the Report or the Funnel that uses the classifier, in its Structure step. With multi-label classification, a piece of feedback gets every theme it mentions, and each theme gets its own sentiment. See Multi-label vs. Single-label Classification.

Example

Take this review of a fitness app: "Since the last update the app keeps logging me out, but the new workout plans are the best I've tried." A content classifier tags it with Login Problems and Workout Plans. With multi-label classification, Login Problems is negative and Workout Plans is positive.

The labels of a custom model can be grouped into categories, such as a Payments category that holds Refunds, Card Payments and Invoices. Feedback is still classified with the labels, and Reports and Spaces show the labels under their categories, so a long list of themes reads like a table of contents. See Organize Labels into Categories.

Dimension Models

A Dimension model answers an analytical question about feedback, such as who is talking or what goes wrong. Where a classifier tags what a piece of feedback is about, a Dimension model asks one research question of every piece of feedback, and its labels are the possible answers. Feedback that says nothing about the question gets no answer.

Every Dimension model is built on one Dimension. You can choose one of the 10 Dimensions in Kimola's list, such as Pain Points ("What goes wrong or creates difficulty?") or Personas ("Who is experiencing this?"), whose question and description are filled in for you to fine-tune. You can also write a question of your own, with a description of what the model should find. Dimension models are custom models created with a prompt: in the Purpose step, you choose a Dimension in place of a classifier.

Dimension models are used in Funnels. A Funnel holds one Dimension model per Dimension next to its content and sentiment classifiers, and every Space that uses the Funnel asks the same questions of every new file, so the answers compare side by side across Spaces and over time. Funnels are available in the Enterprise package. A Report doesn't need a Dimension model: you choose its Dimensions from Kimola's list while you create it. See Add Dimensions to a Funnel.

Note

Funnels, and with them the Dimension models of your Spaces, are available in the Enterprise package.

Example

A hotel group's Funnel might hold a Social Context model with the answers With family, On a business trip and With a partner. The review "Check-in took 40 minutes and the kids were exhausted by the time we reached the room" answers With family.

Extractors

An extractor picks out entities from feedback: the people, organizations, brands, products and locations a text mentions. Where a classifier decides what a piece of feedback is about, an extractor lists what it names.

Kimola's extractor is the Entity Extractor, one of the pretrained models. Use it when your question is which names come up in customer conversations, such as the competitors customers compare you with or the places a complaint comes from. Its page on kimola.com/pretrained-ai-models describes what it recognizes.

Example

In comments under a mobile network operator's social media posts, the Entity Extractor might pick out the rival operators customers compare it with and the towns where they report weak coverage.

Pretrained and Custom Models

Every classifier, Dimension model and extractor is either pretrained or custom. The difference is who defines its labels.

Pretrained models are ready-made by Kimola. There are 15: the Automatic Classifier, the Sentiment Classifier, the Entity Extractor, the Consumer Feedback Classifier, classifiers for e-commerce, mobile apps, mobile games, software, online courses, hospitality (HoReCa), banking, automotive and mobile networks, the News Classifier and the Hate Speech Detector. A new Report starts with the Automatic Classifier and the Sentiment Classifier. See List Pretrained Models.

Custom models use your own labels, so the themes match the products, processes and problems your teams already name. You can create one with a prompt, where you choose the purpose and define the labels, each with a name and a description; with a training dataset, a file with a text column and a label column; or with a Report, from the labels of a Report you have already analyzed. A prompt-based model can also skip defining labels: the AI discovers them from each dataset by answering your question, right before classifying it. Custom models aren't part of the free Starter package. See Custom vs. Pretrained Models.

Note

Custom models are available in the paid packages, not in Starter. See kimola.com/pricing.

Your custom models are listed under MODELS in the main menu, each with its kind, its number of labels and its consistency score, the percentage of semantically consistent pairs in its records (see What is Model Consistency?). To add one, click Create AI Model and choose By Prompt or By Training Set. Pretrained models have no menu item of their own: you choose them in the Classifiers step of a new Report, under PRE-BUILT MODELS, next to your CUSTOM MODELS. Each pretrained model opens a page with its labels and a console to try it. A Funnel takes custom models only.

You can try a model on your own text or on a file before you use it in an analysis. See Test a Custom Model and Test a Pretrained Model.

Every model you add to a Report or a Funnel uses queries: queries = rows × models × Dimensions. See How to Calculate Queries.

In Short

Models turn free text into labels you can count, filter and compare. Each kind answers a different question about the same feedback.

Classifiers tell you what each piece of feedback is about (content classifiers) and how the customer feels (sentiment classifiers).

Dimension models ask one research question of every piece of feedback, such as what goes wrong or who is talking, and give Spaces their Dimensions through a Funnel.

Extractors list the people, organizations, brands, products and places that feedback names.

Pretrained models work right away. Custom models use your own labels and give your organization one language for all its feedback.

Was this article helpful?

Tell us about your thoughts and experiences regarding the article.