Maze

I tested Maze across two product cycles: an unmoderated prototype test on a new pricing page in Figma and a tree-test on a navigation restructure for a SaaS app. The cohort I had in mind was product designers and PMs who need quantitative validation on design decisions before engineering builds anything, distinct from Dovetail which sits downstream and analyzes already-collected qualitative research.

The Figma plugin sync is the standout. I built a flow in Figma, hit the Maze button, and within seconds had a clickable test ready to send to participants with first-click heatmaps, time-on-task, and completion rate captured automatically. The unmoderated tree-test for IA validation surfaced a category mislabel I would have shipped: the dashboard showed 38 percent of testers clicked the wrong primary nav item, which converted directly into a rename in the next sprint. The recruitment panel let me filter by job title and geo and got me 50 product manager respondents within 12 hours at a per-respondent cost that was reasonable for a quick directional test.

What surprised me was how shallow the qualitative depth gets when you actually need it. Open-text questions are supported but the analysis tools are basic compared to Dovetail or even a manual spreadsheet pass: no sentiment clustering, no AI tagging that worked reliably on my test data, no way to export clean themes. The pricing also escalates quickly once you cross from prototype tests into full product testing across multiple flows: the Pro plan limits stack up if you run 4+ studies a month with mixed methods. Some Figma interactions also did not translate cleanly to the Maze player, particularly conditional show/hide states that depended on prior frame interactions, which forced a few tests to be rebuilt as simpler linear flows.

Maze is the right pick if you are validating design decisions before code with quantitative methods like first-click, tree tests, and basic preference tests on a fast cadence. Skip it if your research is primarily qualitative interview-driven (Dovetail), if you need moderated user research with live interviewers (UserTesting or Lookback), or if your tests require complex Figma prototype state that Maze player cannot reliably render.