The July update never quite made it out of the door. June and July ran into each other, and we spent a chunk of both repainting the place. So this one covers the whole summer: a new brand, and sixteen releases across two months. Here's the short version, with links if you want to go deeper.
A new look for MX8 Labs
We've refreshed the MX8 Labs brand: new logo, new palette, refreshed typography. The mark keeps the MX8 you already know, gives it room to breathe, and ties it to what we do: fielding surveys. We also think the colors read better on a chart than the old one did, which was rather the point.
The refreshed MX8 Labs mark.
Bring responses in from anywhere
Two months of work on respondent sources, and the headline is that MX8 Labs no longer has to be the only place your data was collected.
- Third-party imported responses. Ingest results from another survey platform and report on them alongside everything else. Validation summaries tell you clearly what came in, what was excluded and why.
- Twilio text-to-caller. Respondents text in and get called back — a new route into voice surveys that doesn't depend on them answering an unknown number.
- Synthetic profiles. A synthetic respondent source driven by your own prompts, for when you want a directional read before you spend on real audiences.
- Trust all respondents. A single checkbox to turn off fraud detection on any source — useful when you already know exactly who you're talking to.
Setting up a third-party import respondent source.
Learn more: Setting up a Third-party Import respondent source
Exports your analyst will actually accept
Delivery day got shorter. Raw wide Excel exports no longer time out, and there are now uncoded wide Excel and SPSS respondent exports, respondent reconciliation downloads, and the option to include non-complete respondents in wide raw exports. Expanded SPSS fields get compact indexed variable names while keeping their descriptive labels — and long labels stay distinct instead of colliding at SPSS's 256-byte limit.
Multi-sheet Excel workbooks now open on a Contents sheet, with navigation back to it from every tab. Small thing; makes a forty-sheet workbook usable.
The new Contents sheet in a multi-sheet Excel report workbook.
Chat that does the work, in the right place
Chat used to be an excellent advisor that stopped at the suggestion. It can now take actions in the UI — drafting changes to the survey, report or insight you have open, with every edit reviewed and applied by you. In July we scoped it to the active survey, so an edit can't land on the wrong one, and moved the back end onto MX8 Labs' own infrastructure with regionalized deployment domains.
Chat, working against the survey you have open.
Read: Meet Chat: A Research Assistant That Takes Action
Voice surveys that hold a conversation
Voice moved to GPT-Realtime-2.1 with voice-matched holding audio, better interruption handling and recovery from failed turns. Outbound calls can now carry optional call-center background ambience, and inbound call answering is supported. Stale answers, mistimed interruptions and stalled responses are all fixed, and calls no longer time out at the two-hour mark.
Reporting
A run of additions rather than one headline. New aggregations: mean without outliers, selected-count mean for multi-selects including grouped outputs, a histogram-style count of options selected, and uncoded raw response configuration for standard questions. Any respondent metadata used in the survey is now reportable, several questions can share one reporting label, and cross-tab reports can be copied within a survey or project.
Pixel delivery reporting arrived in June, making exposure-source delivery performance easy to review. Caches behave better too: reports recalculate less often, refresh correctly when a quota change closes a field, and no longer stall on weighting errors from datasets that were skipped.
Also new this summer
- Images on ranking questions, and explicit image resolution options for rendered survey images.
- Rebuilt mobile grid layouts for grid select, numeric, rating and this-or-that grids, plus radio buttons on single-select lists.
- Bulk Excel import and export for survey media metadata, with system-managed fields protected.
- Numeric range button questions, and a new location question replacing the old helper.
- Quota editing that copes with min/max-only quotas, zero percentages, variable-backed names and mixed editable groups.
- Faster survey saving, AI model upgrades across the board, and new API documentation and examples for data prep.
From the blog
- The Research Agent Stack: Why Most Platforms Can't Be Automated — Tom, on why the API is the most important feature no one talks about.
- MaxDiff Without the Complexity — Megan, on preference ranking without a stats team.
- Exposure Analysis Belongs in the Survey, Not After It — Tom, on why joining exposure data after fieldwork costs you dose, recency and decay.
- Replacing Your 40K Always-On Program — Tom, on the economics of always-on brand monitoring.
We'd love your feedback!