Create a Report Using a Custom Model

5 mins read - Updated on Nov 15, 2026

You can create a Report with a custom model when you want Kimola to classify your feedback under your own labels instead of labels it discovers from the data. You add your data, choose the custom model in the Classifiers step of the Report wizard, and Kimola labels every row with it, so the Report counts feedback in the words your teams already use.

A custom model is reusable. Once it exists, every new Report can classify its feedback with the same labels, so results from different datasets and different months stay comparable. A Report can use a custom model on its own or next to the Automatic Classifier, the Sentiment Classifier and other pretrained models.

Getting Ready

Custom models are available in the paid packages (Basic, Standard, Business and Enterprise), not in Starter. Sign in to Kimola, open Reports from the main menu, and make sure your account has at least one custom content classifier or sentiment classifier. If it has none yet, create one first with a prompt, with a training dataset or with a Report.

When to Use a Custom Model in a Report

Choose a custom model when the themes of a Report have to mean the same thing every time you read them. The Automatic Classifier discovers themes from each dataset, so two datasets can end up with different themes. A custom model keeps the labels fixed. You can track the same KPIs or topics over time, analyze feedback by business-specific or research-specific categories, and compare Reports without redefining labels.

Example

A hotel group could create a content classifier with the labels Check-in, Room Cleanliness, Breakfast, Noise and Staff. A Report made from guest reviews in March and another made from reviews in June then count the same five themes, so the group can see whether complaints about Noise dropped after a renovation.

Step 1: Add Your Data

Start on the Reports page, where the creation area takes feedback from three sources. The custom model works the same way with each of them.

Upload a dataset. Drag a file into the upload area, or click the area to choose one. Kimola accepts one file in .xls, .xlsx, .csv or .tsv format, up to 30 MB. See Create a Report by Uploading a Custom Dataset.

Paste a link. Paste the link of a product or business page from a supported platform and click Start. To analyze several sources in one Report, use add multiple and put each link on a new line. See Create a Report by Adding Links.

Use an integration. The Integrations row lists Intercom, Zendesk and Google Sheets. See Create a Report Using Integrations.

Step 2: Set Up Your Data

What this step shows depends on where your data comes from. For an uploaded file, an integration or a Google Sheets spreadsheet, the Sheets & Columns step shows a preview of your data and asks you to choose its columns. The Text Column, the column that holds the text to analyze, is required. The Url Column, which holds the web address of each row, and the Date Column, which holds its date and time, are optional. A Date Column lets the Report show how your themes change over time.

For a link, there are no columns to choose. Instead, you choose the size of the Report with the slider under Choose the Report Size, and confirm that you will use the extracted reviews for research purposes only and comply with the terms and copyright rules of the platforms. You can't continue until you agree.

When the step is complete, continue to the Classifiers step.

Step 3: Choose Your Custom Model

The Classifiers step lists the models Kimola can label your feedback with. Kimola's pretrained models, including the Automatic Classifier and the Sentiment Classifier, are offered here next to your own models, which you find under Custom Models. Pick your model from the list, or type part of its name into the filter field to find it faster.

A Report takes custom classifiers of two kinds. A custom content classifier labels feedback with your topics, and its labels become the themes of the Report. A custom sentiment classifier labels feedback as Positive or Negative, judged by the descriptions you wrote for those two labels.

You can choose more than one model. Using a custom model doesn't stop you from using the Automatic Classifier or a pretrained model in the same Report. The Report then shows the results of each classifier, so you can compare your own structured labels with the themes Kimola discovers from the data.

A Report needs a content classifier and a sentiment classifier together to show its best and worst performing themes. Your custom content classifier can be paired with the Sentiment Classifier, or with your own custom sentiment classifier.

Note

A Dimension model can't be added to a Report. A Report takes its Dimensions from Kimola's list in the Dimensions step. Dimension models are used in Funnels, which classify the feedback of Spaces.

Step 4: Set the Structure

The Structure step decides how many labels one piece of feedback can get. Enable Multi-label Classification lets a single piece of feedback receive several labels when it mentions different aspects. Turn it off to group each piece of feedback under one main theme.

With a custom model, multi-label means several of your own labels on one row. A review that says the delivery was late and the packaging was damaged can get both Delivery and Packaging, if your model has those labels. Multi-label gives a more complete view of what each piece of feedback says, and single-label gives clean counts in which every row belongs to exactly one label. See Create a Report Using Multi-Label Classification and Multi-label vs. Single-label Classification.

Step 5: Add Dimensions (Optional)

In the Dimensions step, you can add research questions that Kimola asks of every piece of feedback, such as Motivations, Pain Points, Unmet Needs and Personas. They are grouped as Drivers, Issues, Actors, Context and Comparisons, and they add to the labels of your custom model rather than replacing them. See What is a Dimension? and Add Dimensions to a Report.

Dimensions are optional. Skip the step if you only need your custom model's labels and sentiment.

Tip

The Dimensions you select appear under My List on the left, so you can review and adjust them before you create the Report.

Step 6: Review and Create the Report

The Review step sums up your choices before the analysis starts. It shows the Report Title, the Report Language, where the data comes from (Source, Sources or Dataset) and the Required Query. The title names the Report in your Reports list. The Report Language is the language the Report is written in: English, Spanish, French, German, Portuguese, Dutch or Turkish. Your feedback itself can be in any of the 30+ languages Kimola analyzes. See Language Options in Reports.

Note

Analyzing a Report uses queries: one analyzed row, multiplied by the number of models and the number of Dimensions. Every custom model you add counts as a model, and every Dimension adds to the total. For the queries each package includes, see kimola.com/pricing.

Confirm to start the analysis. A Report from a link first collects the feedback from your links, and every Report then classifies it with your models. Processing time depends on the amount of data. When the Report is ready, Kimola shows "Here we go!" and View Report opens it.

Read the Results

A Report made with a custom model opens on Overview in Reports. Your model's labels appear as the themes of the Report, and the sentiment and the best and worst performing themes appear when the Report has a sentiment classifier and a content classifier. On Content in Reports, you can filter the records by your labels and read the feedback behind each one, as described in Filter Records in Reports. Performance in Reports ranks your themes by their Satisfaction Score and Dissatisfaction Score.

Custom model labels and aspect-based sentiment are shown together, in a structured and filterable format.

If your custom model groups its labels into categories, the Report shows the labels under their categories, and a click on a category lists its labels. You can add or change categories at any time from the Models list. See Organize Labels into Categories.

What Happens Next

To keep your analyses comparable, choose the same custom model whenever you create a Report for the same kind of feedback. Before you use a model in a new Report, you can try it on your own text or on a file with Test a Custom Model. If its labels don't fit your feedback as well as you expected, see Optimize Prompt-Based Custom Models or Optimize Dataset-Based Custom Models.

Was this article helpful?

Tell us about your thoughts and experiences regarding the article.