Changelog

See how Synthetic Users group your content and name the groups. Explore their reasons, compare patterns, and turn the result into a tree test draft.

See whether Synthetic Users can find information in your menu. Follow their routes, inspect wrong turns, and test clearer labels with a follow-up task.

IRIS gets a round of improvements to help you get more out of your research conversations.

Set what you want to learn before generating participants for a concept test, whether you test one concept, compare several or test each separately.

Deploy SU quant module v for N>50 and N<300 users

Synthetic users now draw on improved short term memory, so longer interviews stay coherent even after the conversation grows past previous limits.

Create audiences with a guided builder that also lets you upload an existing audience description document instead of typing free text.

You can upload multi-page brochures and get synthetic user reactions across every page, not just single images.

Have reports in the user testing feature. Prototype here:

Uploaded data is now processed with better evaluation and an anonymization filter, improving accuracy and privacy in your studies.

Include Iris in interviews so users can create custom reports.

Synthetic Users will be able to react to videos, showing where in the video SUs experience emotions and identifying what those emotions are.

Fix audience-crafter subsegmentation (only geographic today → honor gender/other %s), kills the "skews female → 100% female" bug

Deploy report on UX testing studies. Currently disabled.

Allow your LLM of choice to connect with Synthetic Users.

See a clear explanation of how population sizes are calculated, including support for informal geographic areas.

Some clients want the ability to export powerpoint and upload powerpoint templates. Powerpoint export on studies.

Adding the ability to pick numerical audiences from census around the world. Also adding the ability to save audiences.

You can upload images and reference documents in IRIS, giving it more context to draw on when generating responses.

Self-improvement loop of interviews.

Improving quality of the sessions output and adding followup questions to staging.