Practical Guide: Successful Use of AI in Market Research
workshop
| 🗓️ Event | Succeet 2025 |
| 📅 Date | 13 February 2025 |
| 📍 Location | Wiesbaden, Germany |
| 🌐 Language | German |
| 📥 Materials | Slides (PDF) |
My colleague Oliver Tabino and I gave this presentation at the Succeet 2025 conference in Wiesbaden, which is the leading trade fair for market research in Germany.
We presented five case studies of using AI in market research. Four successes, one failure. Then we shared our lessons learned and gave some practical advice on management of AI projects.
Case Studies
The case studies presented are, in order of project size:
- GenAI-based storytelling: Using AI as a sparring partner for marketing
- Topic clustering and labeling tool to analyze answers to open-ended questions in surveys
- LLM-powered social media data pipeline for a client in the pharmaceutical industry
- AI Panel: Virtual respondents for online surveys. This was a research project that ended with the result that current language models are not suitable for this application.
- AspectWise: a data pipeline that uses LLMs to analyze customer reviews and create in-depth reports
Lessons Learned
Our main lessons learned can be summarized as follows:
- Automate your evaluations
- Use prompting for as long as possible before you start with fine-tuning
- View models as exchangeable parts of your pipeline
- Most companies’ market advantage comes from their data, evals and app/presentation layer, rather than a model
- Look at the data and try doing the tasks yourself before automating them
- Projects have a limited innovation budget: innovate on one or two aspects of your project, not all
- “Real artists ship” (Steve Jobs): Focus on one key use case and launch your project, rather than building a complete solution on day one