It's been a busy month: eight releases went out, and we came out of an independent security audit with a SOC 2 Type 1 attestation. A handful of features are worth flagging too. Here's the short version, with links if you want to go deeper.
Your security review just got shorter
MX8 Labs is now SOC 2 Type 1 attested, following an independent audit by Prescient Assurance, a licensed CPA firm. SOC 2 is not a certificate you buy: it is an auditor's written opinion that your security controls are suitably designed and in place, examined against the AICPA's Trust Services Criteria, access control, change management, incident response and vendor management.
Type 1 covers how those controls are designed at a point in time. Type 2 examines how they operate over a period of months, and we are collecting evidence for that now. The practical difference for you: when your security team asks, they can read an independent auditor's report instead of a questionnaire we filled in ourselves.
Read: MX8 Labs achieves SOC 2 Type 1 attestation
Qualitative depth without a separate project
You can now run qualitative text interviews inside a normal survey. A respondent answers one open question, then the AI asks focused follow-ups based on what they've just said, you set the maximum number, and the interview can finish early once the objective has been covered. Interviews resume safely if someone drops out mid-way. Each one comes back as a respondent-level transcript and flows straight into the open-end coding, code-frame and example-verbatim workflows you already use. No separate qual tool, no separate file to reconcile. AI survey programming can create and maintain these interviews directly from a research brief.
Learn more: Running qualitative text interviews · Qualitative text question
Clearer cross-tabs, insights that show their working
Cross-tabs now render as separate per-question tables, and even the largest of them are now quick to scroll through. Your column and row choices persist when you leave and return, and exports follow the same ordering. Two new aggregations arrived, numeric median and selected-count median, for when a skewed distribution or a handful of extreme values would pull the mean away from the typical response. Percentage-based calculations can also use the number of respondents exposed to each individual option rather than everyone who saw the question.
Each question now gets its own table, with columns and rows you can arrange per question.
AI insights now show their working: a numbered evidence reference sits beside each numeric finding, with a References section giving the respondent count, confidence level and weighting behind it, so you can check a claim before you quote it. Reports also refresh reliably after weighting, question and audience changes, and final insights wait until every linked survey is complete.
Learn more: Interpreting crosstabs · Aggregation types · Insights in MX8 Labs
Data out in the shape your analyst wants it
Three new SPSS options. Q-optimized exports open in Q with multi-select questions already recognized as Pick Any, and keep the order options were picked in. SPSS analysis bundles split repeating responses into their own files alongside the main respondent data, with the merge syntax to put them back together and an index explaining what each file holds. And data-and-codebook ZIPs come in wide or stacked, coded or uncoded.
The SPSS download cards, including the new Q-optimized bundle.
Raw Excel and CSV downloads now ship a data map showing which question and answer each column came from, worth a look before you rename anything or hand the file to someone else. Exports also prepare in the background now: a downloads panel keeps track while you carry on working, offers a retry if one fails, and remembers the file type you last chose. Large report exports that used to fail now complete.
Learn more: Downloading data and reports · SPSS export format
More control while you're still in field
You can now change a survey's overall target response count while it's in field, without taking the study down. Twilio Voice, Text and Text-to-Caller sources can now tell you what happened to every call and message you sent. Download a dispatch CSV and you get one row per attempt: who you tried to reach, which number you called from, when, and how it ended. Answering machines, voicemail and fax now come back as no-answer rather than something rosier.
Dispatch CSV, on the field report for a phone respondent source.
Voice campaigns can also be retried against approved per-contact caller identities, and every attempt is logged. Voice surveys now handle calls longer than fifteen minutes. Progress by respondent source now shows when the first and last responses actually came in, in local time, alongside live incidence. On the exposure side, S3 exposure sources now keep a history of every file you send us: how many rows arrived, how many made it in, how many were rejected and why. A later good upload no longer wipes the record of one that went wrong.
Learn more: Fielding reports · Twilio voice respondent sources · Monitoring S3 exposure ingests
Also new this month
- A visual weighting editor for flat, hierarchical, joint and multi-stage schemes built in a table rather than in code, and checked for gaps before you submit.
- Searchable dropdowns for single- and multi-select questions, so a long country or brand list no longer means a page of scrolling.
- Independent MaxDiff randomization of items and of the Least/Most labels, so your existing reports keep working.
- A cleaner survey media editor, with easier bulk import and export of the details attached to each file.
From the blog
- The Squeeze: More Studies, Faster, for Less, and the Way Out — Tom, on the trap agencies are caught in, and why it isn't the team's fault.
- Know Your Non-Converters: The Audience Your Pixel Never Sees — Tom, on the 97% your analytics can't describe.
- Brand Lift Without the Six-Week Wait — Megan, on turning a report card into a steering wheel.
- Keep Your Process. Lose the Busywork. — Megan, on automating the steps you already send away.
Any and all feedback greatly appreciated!