AI for market research
Give it the study folder and it drafts questionnaires and discussion guides, codes the open ends, runs the weighted toplines and crosstabs and builds the wave report. It reads the .sav and CSV exports, transcripts and decks already there.
Tracker-toplines-Sep.xlsxExample| KPI, Top 2 Box | Aug (A) | Sep (B) | Change | Sig. at 95% |
|---|---|---|---|---|
| Aided awareness | 71% | 72% | +1 | No |
| Consideration | 38% | 43% | +5 | B over A |
| Purchase intent | 24% | 25% | +1 | No |
| Value for money | 46% | 44% | −2 | No |
| Base, unweighted | 1,204 | 1,211 | Deff 1.31 |
NoteConsideration is the only KPI that moved more than 3 points, and its 5-point rise is significant at 95% after weighting.
Ask it for
“Write the questionnaire for a 12-minute consideration study among US homeowners, with a screener, skip logic and the quota plan for 1,200 completes.”
Questionnaire-consideration-homeowners.docx“Weight the September wave to the census targets, rebuild the toplines and crosstabs with Top 2 Box and 95% column letters, and flag any KPI move over 3 points.”
Tracker-toplines-Sep.xlsx“Code the open ends on why customers canceled into a code frame, tag every verbatim and count mentions per code.”
Openend-codeframe-churn.xlsx“Check whether the four-segment solution holds under bootstrap resampling and build the typing tool for new respondents.”
Segmentation-typing-tool.xlsx“Run the four Van Westendorp questions from the pricing survey and mark the OPP, IPP and acceptable range for the Standard plan.”
VanWestendorp-PSM.xlsx“Build the Q3 coverage report for Renway with share of voice against Casswick and Ellery, sentiment by theme and message pull-through, and no AVEs.”
Coverage-report-Q3.pptx
Inside the Market Research & Insights mode
Survey design and analysis
Drafts questionnaires with screeners, skip logic and quota plans, and turns .sav and CSV exports into weighted toplines, crosstabs and wave-to-wave trends.
Qualitative research synthesis
Codes interviews, group transcripts and open ends to a code frame, keeps a verbatim behind every theme and reports prevalence, rare codes included.
Customer segmentation
You get a segmentation you can defend, with scaled inputs, a segment count argued for, stability checks under resampling and a typing tool for new respondents.
Pricing and packaging
Builds Van Westendorp curves, Gabor-Granger demand lines and conjoint setups, then reads the price points, elasticity and tiers off the results.
Market sizing and evidence synthesis
Sizes markets bottom-up and top-down in one workbook, reconciles sources that disagree and keeps a source on every input, from expert calls to filings.
Social listening and PR measurement
Reads the monitoring exports and reports share of voice, sentiment and message pull-through. It never converts coverage into ad column inches.
- Office skills
- Excel, Word, PowerPoint, PDF
- Tools
- Read, edit and write files, Search files, Run code, Search and read the web, View images
- Connectors
- SharePoint, Google Drive, Gmail, and additional custom connectors
- House style
- Your fonts, layouts and rules, on every file
How it works
- 1
Load the study folder
Put the .sav or CSV export from the survey platform in the folder, along with the questionnaire, the interview and group transcripts, the verbatims sheet, the client brief and last wave’s deck. Keep them where the team saves study files.
- 2
Name the deliverable
Typical asks are weighted toplines, a code frame with its verbatims attached, pricing curves or the sizing build. It writes Excel files with live formulas, cites the file behind each number and notes what it could not reconcile.
- 3
Review it like a study director
Check the bases, the weights and the column letters the way you would on a junior’s tables. You decide what the study says and what the client sees.
Who it’s for
Market researchers, consumer and customer insights teams, research agency staff, and agency brand strategists and planners.
Questions
Can it apply the weights and run the significance tests correctly?
Yes, when the folder says what to weight to and which test to run, or a past tab book shows the format. It follows each method’s rules, shows the bases it used and names the source file for every figure. You review the tables before they reach a client.
Will it invent sources or numbers in desk research?
No, every statement it writes carries the file or page it came from, and it flags what it could not verify instead of guessing. It also asks before it reads a website for the first time. You still check the citations, because a plausible source that does not support the claim is the classic way desk research goes wrong.
Can we put client data or syndicated data into it?
It depends on the contracts, and that decision belongs to your legal team, the client and the data supplier. Client agreements and panel or syndicated licenses often restrict where their data may be processed, and anonymizing respondents does not satisfy those terms. Where the content goes is set out in the next answer, and whoever signs the contract should read it.
Where does respondent data go when we use the app?
Ultramarine Work changes files only in the project’s folder. When a request needs it, it can read files elsewhere on your computer, never passwords, keys or app data, and you can turn that off in settings. The model runs on Paper Instruments’ servers, so the content it works on is sent there to be processed. Paper Instruments is SOC 2 Type II, and our security and compliance details are on our compliance page.
Do we have to tell clients we used AI in the analysis?
Yes, the ICC/ESOMAR Code expects it. Researchers disclose when AI takes part in analysis, reporting or interpretation and how much human oversight there was. Which wording you use is your call, and the review is yours too, since a person reads every file it touched before anything goes out.
What happens to the minority opinions in our verbatims?
They can get flattened, and the code frame is where you control that risk. The mode reports prevalence across sources, including codes that fire rarely but matter, and it attaches the verbatims so you can read them yourself. Keep small but severe codes visible in the count rather than letting volume decide.