What can text analytics for market research do with your open-ended survey responses, reviews and customer feedback? This free, 90-minute course explains the main methods in plain language, including sentiment analysis, AI coding of open-ends, anonymisation, summarisation and response authenticity, and shows how to choose and evaluate the right tools.

The course is taught by Maurice Gonzenbach, co-founder of AI text analytics platform Caplena, and is free to watch on demand thanks to the support of Caplena. It suits researchers, insights and CX teams who need to analyse text data at scale without a technical background.

What you’ll learn about text analytics for market research

By the end of this course, you’ll be able to:

• Explain the key AI terms used in text analytics: artificial intelligence, machine learning, deep learning and natural language processing
• Choose topic-level over verbatim-level sentiment analysis when you need actionable detail on what people like and dislike
• Pick a sentiment scale you can trust, and avoid continuous scores and emotion detection that invite spurious precision
• Judge how AI handles irony, sarcasm and context in survey and CX feedback, and when it needs extra guidance
• Build a structured codebook with AI: a hierarchical, mutually exclusive and collectively exhaustive list of topics
• Contextualise AI coding by reviewing a few dozen responses, rather than manually coding thousands, to improve results and get a quality score
• Anonymise personal data in open-ended responses using tags or random fakes before sending text to external AI tools
• Use large language models for summaries safely, checking for invented quotes and facts
• Evaluate text analytics tools by testing on your own data, checking language support, agreeing an acceptable level of uncertainty and using F1 rather than raw accuracy

Who this course is for

• Market researchers who analyse open-ended survey responses
• Customer experience and voice-of-the-customer teams running feedback trackers
• Insights managers choosing or reviewing text analytics software
• Agency researchers who code verbatims for clients
• Analysts and newcomers to research looking for free text analytics training

What’s covered in the course

Key AI and NLP terms explained

A short, non-technical guide to how artificial intelligence, machine learning, deep learning and natural language processing relate to each other, and why text analytics improved so quickly from around 2015.

How sentiment analysis works in market research

The difference between verbatim-level and topic-level sentiment, how well AI deals with irony, sarcasm and context, why simple categorical scales work better than continuous scores, and why emotion recognition is rarely reliable enough for quantitative analysis.

How to code open-ended survey responses with AI

A four-step process for topic assignment: identify topics, contextualise the AI, report through dashboards, and generate insights. Includes building a codebook with AI and a human working together, zero-shot categorisation, and why reviewing a sample of responses still matters.

Anonymisation, summarisation and authenticity

How to remove personal data from text, where general-purpose large language models do and don’t help with summaries, and early tools for detecting bot-generated survey responses and fake reviews.

Case study: how FlixBus used text analytics after acquiring Greyhound

After the acquisition, Greyhound’s NPS fell 33 points. By coding detractor comments, FlixBus identified punctuality, refunds, delay communication and customer service as the main issues, and cut dissatisfaction with punctuality from 14% to 10% within four months. The case also shows how it shares insights through annual reports, monthly pulse checks and live dashboards.

How to evaluate text analytics tools

Practical pointers: test on your own data, check language support, decide how much uncertainty your business can tolerate, measure quality with the right metric, and be aware of bias inherited from training data.

Choosing and connecting text analytics software

The trade-offs between user-facing applications, APIs and libraries, why API access matters, and options for moving data between tools, from file uploads and direct integrations to Zapier, n8n and custom code. The course ends with where automation should stop and human review should stay.

About the instructor

Maurice Gonzenbach is co-founder of Caplena, an AI-powered text analytics platform for analysing customer feedback and open-ended responses. An engineer by background, he specialised in numerics, machine learning and natural language processing during his master’s, completed in 2016, and founded Caplena based on technology developed during his studies. Caplena works with large corporations, smaller companies and research agencies.

Frequently asked questions

Is this text analytics course free?
Yes. Introduction to Text Analytics for Research is free to watch on demand, supported by Caplena.

Do I need technical skills to learn text analytics for market research?
No. The course is deliberately non-technical and practical. It explains the key AI terms in plain language and focuses on methods and tools you can apply to survey, CX and feedback data.

How long is the course?
About 90 minutes, including two short question-and-answer sessions with the instructor.

What is the difference between verbatim-level and topic-level sentiment analysis?
Verbatim-level sentiment gives one overall rating to a whole comment. Topic-level (or aspect-level) sentiment rates each topic mentioned, so “Too costly, but the service is great” is negative on price and positive on service. Topic-level analysis is harder to implement but far more actionable.

How many open-ended responses do I need for AI text analytics?
There is no technical minimum, but Maurice suggests that around 300 responses is the point where an AI coding tool starts to save meaningful time and gives a more standardised analysis than coding by hand in Excel.

Is accuracy the best way to measure AI coding quality?
Not always. Most topics appear in only a minority of comments, so a tool can score high accuracy without coding well. Scores such as F1 are better suited to these unbalanced data sets.

Is the course specific to Caplena’s platform?
No. It covers methods and compares tools from a range of vendors.