Hello and welcome to Q4. We're very excited to announce that five new releases went out in September. Most of the month went on two things: the analysis behind conjoint and MaxDiff, and the field report. Here's the short version, with links if you want to go deeper.
Conjoint and MaxDiff, anchored
Conjoint questions have a new respondent experience, and two things you can now ask: an explicit none of these option, and a yes/no purchase confirmation after a choice.
Both feed the analysis. The report editor's Analysis step has Anchored Part Worth Scores and Anchored Attribute Importance, which use those answers to account for whether someone would buy the concept at all, rather than only preferring it to the others on screen. Marginal Purchase Lift needs the purchase follow-ups: it estimates the change in purchase probability when you replace one attribute level and leave the rest as they appeared. Eligible questions get these reports built automatically, with demographic columns.
The Analysis step for a conjoint question.
MaxDiff gets the anchored equivalents, Utility relative to anchor and Probability above anchor, alongside utility scores and simulated share. Simulated share divides preference between the items you tested; the anchored measures say whether an item clears the bar on its own.
Learn more: Choice-based conjoint · Running MaxDiff · Utility and simulated share methodology
The field report, rebuilt
One report, for a single respondent source or for all of a survey's live sources together. The overview covers progress to target, participation, incidence, completion rate, median LOI and daily pace. Below it: breakouts by demographic, quota performance, drop review, a source comparison and technical diagnostics. Set a date window and a report timezone, and download the lot to Excel with the controls you applied.
The overview, for one respondent source.
Median LOI here is active interview time, summed between survey events with each gap capped, not clock time between opening and finishing. On any survey people walk away from and come back to, the two numbers are a long way apart.
You can share a source's report as a read-only link. Name it, choose 7, 30 or 90 days, send the URL. The recipient does not need an account, and the link reaches that source's field report and nothing else. Refresh, reactivate or revoke each link separately.
Named share links, with an expiry.
Field reports also remember the minimum question base you last applied, per report, in the same browser.
Learn more: Field reports · How to set up respondent sources
Cross-tab sorting that sticks
Rows and columns take prioritized sort rules, ascending, descending or manual, and individual values can be hidden. Column rules apply across the report; row rules, hidden rows and widths are per question. Both survive leaving and coming back, and exports use the same order.
Row sorting rules, in priority order.
Respondent Statuses now offers Completes+Partials, and Partials on their own, for reading a study that is still in field. Any individual cross-tab downloads as Excel (Question Tabs), one question per sheet.
Report editing had a broad pass as well: cross-tab and top-box bases, lift reports, video overlays, verbatims and weighting.
Learn more: Reporting with MX8 Labs · Downloading data and reports
Tell the AI how your team reports
The account's Customizations tab has a Get Reports instruction field. Describe the comparisons your team always wants, or how reports should be organised, and generated reports follow it. Instructions are validated before they go live, and you can switch them off without deleting them.
Get Reports instructions, before validation.
Per survey, the Questions tab has breakout switches: pick the questions used for reporting defaults, or restore the AI's selection. Generation is more selective too, choosing a calculation that suits each question and leaving questions with too few responses out of automatic breakouts.
Weighting has a matching rule. A variable now has to be answered by at least 99% of completed respondents to be used, which keeps anything a large part of the sample skipped out of the weights.
Learn more: Customizing AI report generation · Managing questions and reporting labels · Weighted results FAQ
In field, and in test
Configuration changes no longer terminate active respondents. Edit a quota, or anything else respondents never see, and people mid-interview carry on.
Redirects are now set per respondent source: completion and termination delays in seconds, immediate, or no redirect at all.
Redirect delays on a respondent source.
In test, 100 test links in Excel gives each reviewer their own session in one download. The All Testing Report groups attempts by saved survey version, with a chronological drilldown per respondent and timestamps in the report's timezone; clearing test responses no longer wipes it, so a failed test and its fix sit side by side. The test toolbox has an Auto button that submits the current question with its default or simulated answer, for getting down a route quickly.
Surveys can show a progress indicator, with transition messages in the respondent's language.
Learn more: Testing and reviewing your survey · Setting up any sample provider · Configuring how your survey looks
Also new this month
- Respondent sequence in transcripts, supported raw exports and reconciliation downloads, so a transcript matches a row in the data.
- Stable reporting IDs in raw SPSS and wide Excel, with predictable names for expanded questions and readable labels kept in the codebook and data map.
- Descriptive, timestamped download filenames that do not change when the file comes back from cache.
- Chat presents survey-code and data-preparation changes for review before applying them, and keeps its research across specialist steps.
- Multiple exposed dimensions in one survey, including in simulations.
- Display-only prefixes and suffixes on numeric inputs.
- Twilio text sources hold their sender identity once dispatch is queued, report provider failures in enough detail to act on, and have inbound reply routing checked before a source launches or resumes.
From Updates
- Same Survey, Different People: What We Found Testing Chat Against Forms — Megan, on the 40-question study we fielded twice to settle an argument, and the two claims of ours it killed.
- What Happened to Confirmit and Decipher? The Forsta and Qualtrics Timeline — Tom, on the merger trail from Confirmit to Qualtrics, and what to ask if you run studies on one of them.
- Segment x Message x Creative: The Combinatorial Trap of A/B Testing — Megan, on 360 combinations and the traffic to test maybe six.
- From 2 Studies a Year to 40+: Expansion by Utilization — Megan, on what a budget does once a study stops being a commitment.
- MMM Is Back. Here's the Survey Layer It's Missing. — Tom, on a model that tells you what moved and nothing about why.
Always happy to hear your feedback. We'll be at IIEX West the week of October 26th. We'd love to buy you coffee, lunch, drinks or dinner!
Best,
Shannon
805-698-3851