We recently attended the New York edition of the Quirks Conference, where the future of insights was on full display. What were the key takeaways? AI, speed of insights, sample quality, but not as many new ways to solve business problems as we expected.
Here’s our summary and what it means for your insights team.
Building Impact
Insights teams continually seek to engage with their executive teams and key stakeholders. Examples of these successes were shown, including engaging presentations by Sandeep Das on the art of storytelling (my fav…The brain values an experience by the ending. Never end a meeting on a bad note. That could also mean rushing to get through the last slides. Don’t do it. Plan an intentional end.) and Anna Lee providing valuable tips on how to improve report and slide decks to get and keep the attention of executives (my fav - having a clear understanding of audience, solutions, and desired learnings as a guide for reporting decisions).
Advanis values the partnerships that we build with our clients and the deep relationships that we build with both the insights teams and the greater product, marketing, and strategy teams. We’re proud to have sponsored the “Best Client-Side Team” category of the Marketing Research and Insight Excellence Awards (link)
Building Quality
Amidst all the AI advancements on display, the focus on sample quality remains a critical concern, especially for B2B research. Many providers (new and old) are emerging with promises of "verified" samples, it’s essential to scrutinize these claims carefully. Many exhibitors shared their samples with attendees - now the real work of testing their claims begins.
We’re continually testing providers to ensure all projects have reliable trusted data. When evaluating a new provider, we want to know if they can offer unique sources of participants, do they compensate them fairly for their participation, and do they have the tools needed to weed out bad actors and other forms of sampling fraud? Few sources of sample meet our standards - we don't compromise on sample quality.
Building Speed
Generative AI is giving rise to new ways to do research. Insights teams and vendor partners are working together to test the best ways to take advantage of this new technology, while maintaining rigorous research practices that underpin insights. One clear benefit is the speed of research - it is only getting faster!
Synthetic Data
Synthetic data is moving from a cautious concept to a viable tool for market research. Companies like T-Mobile are now successfully augmenting collected data with modeled synthetic data to confidently speak to a wider range of markets. When speaking with Antoinette Sandy from T-Mobile, she shared that in initial testing, they didn’t have confidence in the data, but that in the span of less than a year, they improved so much that they gained confidence quickly.
A contrasting (caution) regarding building fully synthetic samples came from Verasight (who work with academic institutions). They noted that they are concerned that synthetic data models appear to work at aggregate, but fail at the segment level. The error rates at the segment level cancel each other out, making the aggregate prediction appear stronger than it really is.
We’re excited to build statistically reliable models of data collection and analysis that help increase the speed of research - faster decision making is good for businesses and governments.
Qual at Scale & The AI Moderator
AI has become a valuable tool for qualitative research, with AI-powered summarization and thematic analysis, and AI moderation of online interviews. These tools have quickly become a valuable tool in the qualitative toolkit. There is a debate between chat-bot, AI Avatar and Karaoke style of AI moderation. While participants still value human connection, they appreciate that AI can offer a judgment-free space to share their opinions.
Clients need to ask their vendors what they are doing to ensure a great user experience for participants, how they are protecting data and how they are training their teams to use AI summarization carefully, avoiding erroneous conclusions that hallucinating models can still produce.
Here Comes Agentic AI
The recent announcement of ChatGPTs Agent has put a focus on the potential for agentic tools to support researchers and repetitive processes. We saw presentations that touted the use of digital twins (and others that urged caution still), as well as presentations about the use of Agents to support researchers in their day to day tasks.
One thing that is clear - as the way we do research is molded by AI, curiosity and business knowledge will win the day.