3 product updates | Sep 07, 2025

June was about removing friction from the research process. Here's what we shipped:

1. One-click participant recruitment

Recruiting participants just got ridiculously simple. Describe who you want to talk to in plain English and we'll handle the rest. "I need 20 people who manage remote teams and use project management software" becomes a perfectly targeted recruitment campaign.

The AI understands complex criteria and automatically translates them into screening questions. It even suggests similar audiences you might not have considered. One product team discovered that IT managers gave better feedback about their dev tools than actual developers.

Recruitment starts immediately after you confirm. No more building complex screeners or waiting for approval. Studies that used to take days to set up now launch in minutes.

2. Smart insight tagging

Every insight generated by AI Analysis now gets automatically tagged with topics, themes, severity, affected user segments, and related features. But here's the clever bit: tags are semantic, not just keywords.

An insight about "users getting lost during setup" might be tagged with onboarding, confusion, first-time experience, and activation. These tags connect insights across studies, building a knowledge graph of your user feedback.

Search for any tag and see every related insight across all your studies. Finally, a way to answer "What do we know about checkout?" without reading through dozens of reports.

The system also learns your tagging preferences. Manually adjust a few tags and the AI adapts to match your taxonomy.

3. Response quality predictions

Before you launch a study, we now predict response quality based on your study design. The AI analyzes question complexity, study length, participant criteria, and incentive amounts to forecast completion rates, response depth, and likely dropout points.

It's surprisingly accurate. We're seeing 85% accuracy on completion rate predictions and 90% accuracy on identifying problematic questions.

The predictions come with specific suggestions. "Question 7 is likely to cause dropouts. Consider simplifying or moving it earlier when participants are fresh." Or "This incentive amount typically yields rushed responses for 45-minute studies."

One team avoided a failed study when predictions showed their technical screening questions would eliminate 95% of their panel. They adjusted before launching and got perfect participants.

Small but useful: we also added CSV import for bulk participant upload, customizable study URLs for branded experiences, and a duplicate detection system that flags if you're about to run a study similar to one you've already done.

June felt like a maintenance month but these improvements are already saving teams hours per week.

‍

Author

Axel Thomson

Founder

Author

Continue reading

Understand your customers with AI-driven research