No more struggling to create a codebook with unbiased topics that don't overlap nor leaves text data uncategorized. With Caplena, you can upload your data in one click and get a hierarchical codebook including descriptions and frequency for each topic.

Whether you upload 100 or more than 100,000 rows, our proprietary AI uses LLMs to generate a codebook that's statistically relevant, and "MECE" by design — with topics that are Mutually Exclusive and Collectively Exhaustive to truly capture your feedback data.

When manual analysis and other tools lack the flexibility to make changes, Caplena's interactive UI makes it easy to adapt your topic collection to your liking.

You can add new topics with ease, change hierarchy, import your existing codebook, start topic discovery from scratch, or do a mix of both — and keep an audit trail. Advanced text analysis finally becomes enjoyable.

Caplena helps you fine-tune your codebook with the power of Large Language Models. Designed to help you go more or less into details interactively, our built-in LLM-based tools are here to assist you in performing advanced text analysis. Caplena helps you spot potential duplicate topics, rare topics, or explore new ones.

Have a specific topic in mind and want to check for frequency? Use prompts to explore and adjust your topic collection to meet your research objectives.

When analyzing tracking studies, topics change over time, and it’s crucial to identify trends. With Caplena, you can continuously monitor for new topics. Our Topic Assistant will surface them for you, while you can set up Alerts to be notified as soon as new topics are detected and get actionable insights.

Straight out of the box, our industry-leading AI, powered by LLMs assigns topics as accurately as a human would with manual analysis (AI quality score of >70) for over 75% of projects.

Just like briefing elements, Caplena's AI takes the context of the project, question formulation, topics, and column descriptions into consideration when making topic assignment. This approach to text analytics ensures best-in-class coding while taking into account your business context and research objectives.

When coding large volumes of open ends with AI, you want to ensure that assignments are reliable. Caplena offers you a unique quality assurance and fine-tuning process so you can trust your results.

An AI quality score indicates how well the AI understood your data and how reliable the assignments are. You can still fine-tune the topic assignment by adjusting some of the auto-generated topic descriptions or reviewing a handful of rows for quality assurance, then trigger an AI update.

This traceability gives you peace of mind that the results you will share with your stakeholders or clients will actually be relevant, and defensible.

A verbatim often contains multiple topics, and sentiments.

With Caplena, you can go beyond sentiment at verbatim level to capture nuances at topic level, and quantify them at scale at no extra cost. This gives you a strong foundation to create reliable driver analyses, correlations, and trend visualizations.

As a Swiss-based company, understanding multiple languages is part of our culture. Our AI natively supports more than 100 languages.

Ensure consistent text analytics and comparable results across different languages, countries and regions, even within the same project.

Have product names, locations, company or people's names hiding in your open ends, or in your metadata? With Caplena's Smart Columns, extracting them becomes easy.

You can then use these values for slicing & dicing in your reports or creating filters for segmented views, and discover precise, actionable insights.

When analyzing customer or employee responses, you want to make sure you don't show unsuitable content to your stakeholders or clients. Caplena helps you flag and remove profanity from text data, but also correct these pesky spelling mistakes at scale.

Your respondents' personal data must be protected. With Caplena, you can automatically anonymize text data on import so that no personally identifiable information (PII) are inadvertently shared, in compliance with the General Data Protection Regulation (GDPR).

Caplena uses large language models (LLMs), not keyword rules.

To assign topics, the AI reads each response together with your project context:

• project and question descriptions,

• topic and category names,

• a description of each topic.

It assigns topics and topic-level sentiment, and every result is explainable: it highlights the phrase that justified each assignment and shows how confident it is.

You steer it by editing topic descriptions in plain language, then re-running the AI on the affected responses. Results are typically human-level after the first run, so a spot check of 10–15 responses is usually enough.

Each run gets an AI Quality Score from 0 to 100 that measures how consistently the AI codes your data. A score of 70–85 is human level: two people coding the same data typically score 70–85. Results are also repeatable: two runs on the same data agree on about 92% of topic assignments. See how we test repeatability.

We use frontier models from Google Cloud and Microsoft Azure, running in their EU data centers. Your data is never used to train them. See more on our Trust Center.

Unlike a black box, Caplena provides you with an AI quality score - a F1 score - measuring how confident the AI is in assigning topics and sentiments to your data. This score guides you to the topics that can be verified first, to achieve a greater score.

As you fine-tune the results, Caplena learns from you and applies your changes to the entire data set.
You can stop fine-tuning once you’re satisfied with the classification.

This process ensures you have a strong text analytics foundation and most importantly, trust in your results.

Whether you are defining a list of topics to track or looking to enhance your codebook, knowing how many topics to aim for is crucial. Our recommendation is to start from
between 30 to 60 topics.

This is exactly the number of topics and sub-topics Caplena will generate based on your data. While you cannot ask Caplena to generate an exact number of topics, you can import an existing codebook and either add or remove topics as you fine-tune your topic collection.

Based on 100,000+ projects that we have analyzed on our platform, we see that about 90% of the projects have between 10 and 100 topics (and then we have a long tail of up to 300+ topics for extremely elaborate cases, like Lufthansa).

AI systems have dramatically improved in their ability to detect irony and sarcasm. Like humans, however, they rely on indications of these linguistic cues in the text — such as exaggerations, emojis or capitalization. While they may not yet be quite as good as humans, they correctly catch the vast majority of cases. One advantage over humans is that they don’t tend to over-interpret, whereas humans often read too much between the lines, which can also skew results.

Caplena supports Excel (xls, xlsx), CSV, and SPSS (sav) files with hundreds of columns and hundreds of thousands of rows. JSON files are also supported via the API, or you can connect your systems through one of our ready-made integrations. Alternatively, let us gather online reviews for you directly from one of the supported sites.

No. Your data is not used to train Caplena's AI or any third-party model. For AI features, we use frontier models from Google Cloud and Microsoft Azure, running in their EU data centers. Caplena may use usage data (how the product is used) to improve our services, benchmark performance and test new AI algorithms. See our Trust Center for all subprocessors and where they process data.