Market research has come a long way since its inception. The rise of technology has made it easier for companies to gather insights about their target audience and make informed business decisions. One technology that is poised to have a significant impact on market research is ChatGPT, an AI-powered language model developed by OpenAI.
ChatGPT is designed to understand human language and generate responses based on its training. This makes it an ideal tool for market research as it can mimic human conversation, providing valuable insights into consumer behaviour and preferences.
One of the key advantages of using ChatGPT in market research is its ability to handle large volumes of data. ChatGPT can analyse vast amounts of information in a matter of seconds, making it much faster and more efficient than traditional market research methods. This can lead to more accurate insights and improved decision-making for companies.
Another advantage of ChatGPT is its ability to analyse unstructured data. Traditional market research relies on structured data such as surveys and focus groups, but ChatGPT can analyse unstructured data from sources such as social media and customer reviews. This can provide a more comprehensive view of consumer behaviour and preferences, leading to more accurate insights.
ChatGPT also offers the ability to personalise market research. By mimicking human conversation, ChatGPT can engage with consumers in a way that feels natural, leading to more accurate and reliable responses. This can provide a more in-depth understanding of consumer behaviour and preferences, and help companies tailor their products and services to meet their needs.
In conclusion, ChatGPT has the potential to revolutionise the market research industry. Its ability to handle large volumes of data, analyse unstructured data, and personalise research make it an ideal tool for companies looking to gather accurate and reliable insights into consumer behaviour and preferences. As AI continues to evolve and advance, we can expect to see ChatGPT play an increasingly important role in market research and other industries.
Integrating ChatGPT into the 2CV toolkit
If you haven’t guessed it already, to this point, this entire blog post has been written by ChatGPT! At 2CV, we’ve been at the forefront of this AI ‘revolution’ pushing the tech to its limits and working out what it can do, and where we can use it to improve the processes and insights we generate.
Here are some of the big and small things where we’ve found benefits from integrating it into our toolkit:
- Analysing/ finding themes in unstructured data. This could be qualitative transcripts or quantitative verbatim responses. However, its ability to do this right now remains very limited and misses much of the nuance an experienced researcher would pull from the same data set
- Translating open-ended responses or questions.
- Aiding questionnaire/discussion guide development. For example, by summarising a list of motivations or behaviours it could help researchers develop answer code lists or discussion points
- Analysing and summarising vast amounts of unstructured social conversations
- Generating catchy or alliterative titles or headlines
- Generating ideas for segments names
- Rephasing or simplifying the language in reports/proposals
- It will help the advancement of Conversational AI (chatbots), an emerging technology we’ve already found helps get qualitative insights at scale
- Asking it for ideas on just about anything
- Helping with the creation of blog posts 😉
Exercising caution while using ChatGPT
What’s clear is that whilst this technology is a powerful tool, it’s not a replacement for human experience and creativity. Its intelligence may be artificial but its training is based on words written by people and unfortunately it picks up and displays some of our biases - something us researchers must be aware of and monitor in all of our work. In another parallel to traditional market research, ChatGPT is still tied to the old adage garbage in, garbage out. Give it the wrong prompt and it will produce the wrong results, just like if we used a poorlyworded question in a traditional online survey.
When you think about the ‘prompt’ being a huge amount of unstructured data – there is a real risk of bias becoming a bigger problem. When the tool is being used to summarise or otherwise synthesise aggregated data, there are considerable issues in having the model fairly represent a diversity of opinions. For example, hallucinating things which were not said, or simply generating descriptions which represent the majority of text it has been trained on, but not necessarily fairly representing the nuances of the respondent verbatims.
When it comes to generating questions, a key flaw of using ChatGPT is that it has no empathy – it doesn’t have any understanding of how questions it asks may be received. It tends to assume, for example, that people can easily and simply articulate how they feel– which is often difficult to express. It also holds no understanding of how ‘big’ of a question it’s asking (e.g. it would treat the questions ‘what is your dog’s name’ and ‘what is your financial situation?’ equally). Because of this – the questions or response options that ChatGPT generates require oversight to ensure it’s asking, not just the right questions, but the appropriate questions.
ChatGPT is a fantastic multi-purpose research tool but, just like any tool, researchers need to know when and how to use it to get the best results.
