Maze
Continuous product discovery: unmoderated testing and surveys on prototypes and live products
Maze is a user research platform for product teams, built to run unmoderated studies at speed: prototype and live-site usability tests, surveys, card sorts, tree tests, five-second tests, and interview studies, with automatic metrics and an AI-assisted analysis layer. It integrates directly with design tools so a Figma prototype can be tested with real participants within an hour of being finished.
Overview
Maze was built around a specific frustration: research was too slow to influence decisions that were already being made. Its answer was unmoderated testing that a product manager or designer could launch alone, with results computed automatically rather than assembled by a researcher afterwards. That positioning, research as a continuous input rather than a quarterly project, is what the company now calls continuous product discovery.
The workflow starts in the design tool. A Figma prototype connects directly, tasks are defined against its frames, and participants complete them while Maze records paths, misclicks, time on task, and completion rates. The output is a report with quantitative metrics alongside heatmaps of where people clicked, which is what makes an unmoderated study persuasive to stakeholders who distrust anecdote.
Beyond prototype testing the platform covers live website testing, surveys, and the classical information architecture methods, plus a participant panel for recruiting and an interview product for moderated work. AI features increasingly handle the analysis burden, clustering open-text responses and summarizing sessions. The trade-offs are price, which is aimed at funded product teams rather than solo operators, and the inherent limitation of unmoderated research: you get behavior and answers, but no follow-up question when someone does something surprising.
Best for
Product designers and product managers at software companies who want to validate prototypes and concepts continuously, with quantitative reports polished enough to circulate to stakeholders without editing.
Not the right fit for
- Teams whose main need is analysis of existing site traffic, which is behavior analytics rather than research.
- Solo founders and very small teams on tight budgets, where the entry price is a real barrier.
- Deep qualitative research programs requiring extensive moderated interviewing as the primary method.
- A/B testing and conversion experimentation, which Maze does not perform.
- Organizations needing a full research repository with taxonomy and governance across many teams.
How it works
- 1
You build a study from a template or blank, choosing a method: prototype test, live website test, survey, card sort, tree test, five-second test, or interview study. Prototypes connect from Figma and similar tools so no separate upload or conversion step is required.
- 2
Tasks are defined with a goal and, for prototype tests, an expected path. Follow-up questions can be attached after each task, capturing why someone struggled while the experience is fresh rather than at the end of the session.
- 3
Participants arrive from your own audience via a link, from Maze's panel with demographic and professional targeting, or from an in-product widget that recruits real users. Screeners qualify people before they consume study slots.
- 4
Results compute automatically: completion and misclick rates, time on task, heatmaps per screen, path diagrams showing where people diverged, and clustered open-text themes. Reports are shareable as live links, which is how findings reach stakeholders who will not open a research repository.
Feature breakdown
20 features in 4 modulesPrototype and product testing
Testing designs before they cost engineering time.- Figma prototype testing
- Direct connection to prototypes with tasks defined against frames, no export or rebuild required.
- Live website testing
- Task-based studies against a real site, so shipped experiences can be measured the same way as designs.
- Path and misclick analysis
- Diagrams of the routes participants took and where they clicked something that was not the intended target.
- Task metrics
- Completion rate, time on task, and directness computed automatically for every task in the study.
- Screen heatmaps
- Aggregate click distributions per prototype screen, showing where attention landed rather than where it was intended.
Research methods
The rest of the toolkit in the same account.- Surveys
- Standalone or embedded questionnaires with logic, used for attitudinal data alongside behavioral tasks.
- Card sorting and tree testing
- Information architecture methods with grouping and findability analysis for navigation work.
- Five-second testing
- First-impression tests measuring what people take away from a screen at a glance.
- Interview studies
- Moderated sessions with scheduling, recording, transcription, and note-taking for qualitative depth.
- Study templates
- A library of pre-built studies for common questions, which is often what gets a first study launched at all.
Participants and reach
Solving the recruiting problem that stalls most research.- Participant panel
- Recruit by demographics, location, and professional criteria, charged per participant on top of the subscription.
- In-product recruitment
- Widgets that invite your own users into studies, producing participants who actually match your market.
- Own-audience links
- Share a study with your list, community, or customers at no participant cost.
- Screener questions
- Disqualify unsuitable participants before they consume paid slots.
- Incentive management
- Panel incentives handled by the platform rather than administered manually.
Analysis and collaboration
Turning sessions into something colleagues will read.- AI analysis and clustering
- Open-text responses grouped into themes and sessions summarized automatically, reducing the manual coding burden.
- Live shareable reports
- Results published as a link that updates as responses arrive, rather than a slide deck assembled at the end.
- Video clips and highlights
- Cut and annotate moments from sessions for use in readouts and stakeholder discussions.
- Integrations
- Connections to Figma, Slack, Jira, Notion, Productboard, and similar tools so findings land in existing workflows.
- Team workspaces
- Shared study libraries with permissions, so research accumulates rather than living in individual accounts.
Use cases
4 documentedProduct designer validating a redesign
A new checkout flow is designed in Figma and the team is about to commit two sprints to building it.
An unmoderated prototype test with 40 participants returns completion rates and misclick heatmaps in two days, exposing a step that fails for a third of users before any code is written.
Product manager choosing between two concepts
Leadership is split between two directions and the debate has stalled on preference.
A preference and comprehension study produces quantitative data plus verbatim reasoning, and the decision is made in a meeting rather than deferred again.
Design team improving navigation
Support tickets suggest customers cannot find account settings and billing.
A tree test measures findability across the current and proposed structures, and the improvement is demonstrable rather than asserted.
Startup with no research function
Needs evidence for product decisions but cannot hire a researcher.
Templates and automatic analysis let the product team run studies themselves, with the panel handling recruitment they would otherwise have no way to arrange.
Pricing
from Free tier for limited studies; paid plans commonly from around $99 per monthSeat and volume-based subscription with a free tier for occasional use. Paid plans scale by studies, responses, and seats; panel participants are charged separately per person. Annual billing is discounted and enterprise agreements are quoted.
| Plan | Price | Includes |
|---|---|---|
| Free | $0 per month |
|
| Professional | From about $99 per month |
|
| Organization | Quoted annual |
|
Billing notes
- Panel recruitment is billed per participant on top of the plan, and for a study of any size that cost can exceed the subscription.
- Recruiting from your own users or customer list avoids panel charges entirely and usually produces better data.
- Response and study limits are the practical constraint on lower tiers rather than feature gating.
- Annual billing is discounted; prices as published August 2026 and subject to change.
- Enterprise pricing is seat-driven, which suits teams where many people run studies and less so where one person runs many.
Value assessment: Maze converts research from a scheduled project into something a designer does on a Tuesday, and the time saved before a bad build is the return that matters. For a funded product team, one avoided sprint pays for a year of the subscription. For a small team or solo operator, the entry price plus panel costs is steep next to cheaper suites that cover the same methods with less polish, and the honest comparison is against those rather than against not doing research at all.
Strengths & limitations
Strengths
- Direct Figma integration makes testing a design as easy as sharing it.
- Automatic quantitative metrics give unmoderated studies stakeholder credibility.
- Reports are genuinely presentable, which matters more than researchers like to admit.
- Broad method coverage including IA testing, surveys, and moderated interviews.
- Panel and in-product recruitment remove the usual reason studies stall.
- AI analysis materially reduces the manual work of coding open-text responses.
Limitations
- Unmoderated testing cannot ask the follow-up question when a participant does something unexpected.
- Pricing is aimed at funded product teams, with panel costs on top of the subscription.
- No A/B testing or conversion experimentation on live traffic.
- Behavior analytics on an existing site is shallow compared with dedicated replay tools.
- Prototype testing inherits prototype limitations: participants behave differently when nothing is real.
- Not a research repository, so accumulated knowledge management needs another tool at scale.
Head-to-head comparisons
4 alternativesMaze vs UXtweak
from Free for small studies; paid plans from roughly $80 per monthThe direct comparison. UXtweak covers a similar method range at a lower price and with strong classical IA tooling, while Maze is more polished, integrates better with design workflows, and produces reports that circulate well internally. Budget-constrained teams and researchers who value method breadth choose UXtweak; product teams who need findings to persuade stakeholders quickly choose Maze.
Full Maze vs UXtweak comparisonMaze vs Hotjar
from Free (35 daily sessions on Observe Basic); paid Observe plans from roughly $32 per month billed annuallyComplementary. Hotjar watches real traffic on a live site and asks it questions; Maze recruits participants and gives them tasks, including on things that do not exist yet. A product team usually wants both eventually, with Hotjar answering what is happening now and Maze answering whether the proposed replacement is any better.
Full Maze vs Hotjar comparisonMaze vs Typeform
from $0 (free, 10 responses a month), then $29 per month billed annually for BasicOverlap only on surveys, and the intent differs. Typeform builds beautiful customer-facing forms for marketing and data collection at scale. Maze's surveys are research instruments sitting alongside usability tasks and IA studies. Use Typeform when the form is the product experience; use Maze when the survey is one part of a study you intend to act on.
Full Maze vs Typeform comparisonMaze vs Attention Insight
from From roughly $19 per month for a small monthly analysis allowanceBoth answer design questions before launch, but on different evidence. Maze recruits real participants to attempt tasks on a prototype and reports completion rates, misclicks, and time on task. Attention Insight runs no participants at all: it predicts first-glance visual salience from a model trained on eye-tracking data and returns a heatmap in seconds, from roughly $19 a month. Use Attention Insight as a cheap composition check on a hero section or an ad creative; use Maze when the question is whether people can actually finish the flow.
Full Maze vs Attention Insight comparisonImplementation & onboarding
- Setup time
- A first prototype study can be built and launched within an hour, particularly from a template. Live website testing and in-product recruitment require a script and slightly more setup.
- Learning curve
- Low for launching studies, moderate for designing them well. The common failure is writing leading tasks, which produces confident but useless data, and templates only partly protect against it.
- Onboarding
- Self-serve with strong documentation, templates, and a substantial library of research education. Organization plans include onboarding and success support.
- Migration notes
- Studies and reports export, but past research from other platforms is generally retained as documents rather than migrated. Because studies are cheap to re-run, teams switching often simply repeat key studies on the new platform rather than porting history.
Platform, API & security
- Platforms
- Web applicationFigma and prototyping tool integrationsIn-product recruitment widgetMobile web testing
- API
- API access on higher tiers for study and response data, plus native integrations with design, workflow, and product management tools.
- Compliance
- GDPRCCPASOC 2 Type II
- Data residency
- EU and US processing options depending on plan.
- SSO
- SAML single sign-on on organization plans.
- Security notes
- Participants consent explicitly as part of study enrollment, which simplifies the privacy position relative to passive session recording; sensitive prototypes can be restricted by access controls.
Support & resources
- Channels
- Email supportIn-app chatDedicated support on higher tiers
- Documentation
- Clear documentation plus an extensive research education library that has become a reference point for product discovery practice.
- Community
- Large following in product design and management communities, with regular events, published benchmarks, and an active user base.
Company
- Founded
- 2018
- Headquarters
- Remote (incorporated in the United States, European roots)
- Ownership
- Private, venture-backed
- Founders
- Jonathan Widawski, Thomas Mary
- Employees
- ~150 (est. 2026)
- Funding
- Raised venture funding including a Series B round.
Timeline
- 2018Founded to make prototype testing fast enough to fit inside a design sprint.
- 2020Figma integration and automatic metrics establish it as the default unmoderated testing tool for product teams.
- 2021Raises a Series B and expands beyond prototypes into live website testing and surveys.
- 2023Adds interview studies and a participant panel, covering recruitment as well as tooling.
- 2025AI analysis features reduce the manual work of clustering and summarizing qualitative responses.
- 2026Positioned as the continuous product discovery platform for design and product teams.
Integrations
- Figma
- Slack
- Jira
- Notion
- Productboard
- Zapier
- Google Analytics
- Miro
Frequently asked questions
10 questionsWhat is Maze?
Maze is a user research platform for unmoderated and moderated studies: prototype and live website usability tests, surveys, card sorts, tree tests, five-second tests, and interviews. It computes task metrics automatically and integrates with Figma so a design can be tested with real participants shortly after it is finished.
How much does Maze cost?
There is a free tier with limited studies and responses. Paid plans commonly start around $99 per month and scale with volume and seats, with organization agreements quoted. Panel participants are charged separately per person, so budget the subscription and recruitment costs independently.
Does Maze work with Figma?
Yes, and this is its signature capability. Prototypes connect directly, tasks are defined against frames, and results include path diagrams and per-screen click heatmaps, so a design can be validated without exporting anything or rebuilding it elsewhere.
What is unmoderated testing and what are its limits?
Participants complete tasks alone, on their own time, with no researcher present, which is why studies can run overnight at scale. The limitation is that nobody can ask a follow-up question when something surprising happens, so unmoderated methods are excellent at measuring what and where and weaker at explaining why.
Maze vs UXtweak: which is better?
UXtweak is cheaper and covers a similar or wider method range, with particularly strong information architecture tooling. Maze is more polished, integrates better with design tools, and produces reports that persuade stakeholders with less editing. Choose on budget and audience: internal researchers often prefer UXtweak, cross-functional product teams usually prefer Maze.
Can Maze recruit participants for me?
Yes, through its own panel with demographic and professional targeting, charged per participant. You can also recruit from your own users through an in-product widget or share a study link with your customer list, both of which avoid panel costs and usually yield more representative participants.
Does Maze do A/B testing?
No. It tests designs and concepts with recruited participants rather than splitting live traffic. Validating a change on real users requires an experimentation platform such as VWO or Convert; the two are complementary stages of the same process.
How many participants does an unmoderated study need?
For qualitative discovery, five to eight participants surfaces most major usability problems. For quantitative comparisons such as completion rates between two designs, you need considerably more, typically 30 to 50 per variant, because you are comparing proportions rather than spotting obvious breakage.
Is Maze suitable for a startup with no researcher?
It is designed for exactly that case, with templates, automatic analysis, and recruitment handled in-platform. The remaining risk is study design: badly worded tasks produce confident wrong answers, so read the guidance on leading questions before running anything that will drive a decision.
Can I test a live product rather than a prototype?
Yes. Live website testing runs task-based studies against a real site, and in-product recruitment can invite actual users into studies. That combination lets a team measure the shipped experience with the same metrics used to evaluate the prototype it replaced.
Editorial verdict
Maze's contribution is speed: it made research fast enough to influence decisions that were already in motion, which is the difference between a research function that shapes a product and one that documents it afterwards. The Figma integration, automatic metrics, and shareable reports together mean a designer can settle an argument with data in two days rather than deferring it. The limits are structural rather than fixable: unmoderated studies cannot probe, prototypes are not products, and panel costs stack on top of a subscription already priced for funded teams. For product organizations that will actually run studies weekly it is excellent value, and for everyone else a cheaper suite covers the same methods with less polish.
Written by the SaaSTracker editorial team. Awards, when shown, are judged against the published criteria in our methodology.