Welcome to May!
It's been a banner month on the platform. April brought five releases and a handful of features worth flagging. Here's the short version, with links if you want to go deeper: cross-tab charts, fine-grained permissions, voice surveys, and a couple of new ways to lean on Claude or ChatGPT.
See your cross-tabs as charts
You can now flip any cross-tab from a table into a chart view: 100% stacked bars for single-select questions, stacked columns for multi-select. Same variable, series, and filter controls as the table view, plus a one-click image download for dropping straight into a deck. Useful for the "show me the shape, not the decimals" moments.
Brand choice by annual income, rendered as 100% stacked bars in the new chart view.
Learn more: Interpreting cross-tabs in MX8
Fine-grained roles and permissions
The account section has been rebuilt around roles. You can now decide exactly who in your team can edit surveys, manage respondents, see exports, and change billing, all configurable per role rather than a blanket admin/user split. If you've been holding off on inviting the wider team, now is a good time.
Learn more: Account roles and permissions
Translate any survey, control how it looks
Two related additions: you can now upload translation files directly in the translation tab (no more round-tripping through email), and survey themes now include controls for the icon and background imagery shown to respondents. Multilingual studies in your brand, without the back-and-forth.
Pick an icon and background imagery for your survey from the new theme controls.
Learn more: Translation upload · Survey theme controls
Voice surveys, now production-ready
Voice calling got a substantial reliability pass this month: answer-capture retries, confidence gating, and turn-handling were all hardened. Combined with voice support on any Twilio number that shipped at the start of April, it's a good moment to try a voice study if you've been on the fence. Megan wrote about why we think the technology is finally there.
Read: Voice Surveys Are Here — Talk to Your Respondents (Literally)
QA your survey with an LLM
You can now export a survey's respondent transcripts as Markdown, hand them to Claude or ChatGPT, and find logic problems, confusing question wording, or fraudulent patterns in minutes rather than hours. Tom wrote a how-to with the exact prompts he uses. Probably the fastest way to catch survey issues we've seen.
Read: How to QA Your Survey Using Claude or ChatGPT
Also new this month
- Live incidence rate is now reported per respondent source, so you can see which sources are actually delivering qualifies.
- You can now mark poor-quality respondents while a survey is still in field, not just after.
- Insights now generate automatically when you close a survey — no more manual refresh after the last respondent comes in.
From the blog
- AI Survey Fraud Is Getting Harder to Catch — Megan, on why detection has become an arms race.
- Build a Research Deck with Claude and MX8 Labs — Tom, on stitching the MX8 MCP connector to your PowerPoint template.
- When the Research Budget Doesn't Grow — Megan, on how teams are being asked to do more with less.
Please let me know if you have any questions or suggestions.