Features & methods.

Cauliflower imports SPSS, Excel and CSV files. A preview shows the first rows. For CSV files, you can choose the delimiter; for the table structure, you define the header row and where the data starts.

Cauliflower is suitable for market research studies using structured survey data and open-ended responses. These include tracking studies, ad tracking, concept and claim tests, ad hoc studies, product tests, brand and image studies, usage and attitude studies, and customer satisfaction and NPS surveys. You can also analyse text outside conventional surveys, such as chatbot conversations, app reviews, Google Maps reviews and customer feedback. Explore topics, ratings and relationships, or ask targeted questions about the text using classification questions.

The software identifies related questions, question types and matrix questions using variable-naming patterns and semantic understanding of survey content. It then proposes a study structure. Before calculation, you review this structure and can merge or separate groups and add question texts, short labels and brand names.

A guided setup takes you through each step: name your job, upload data and review the detected structure. If needed, restrict the analysis to selected questions and columns. You can configure filters, waves and weighting; countries are set afterwards in the dashboard. For text analysis, select text columns and the analysis type; their question base can be restricted using existing routing or split-sample variables. We can agree the setup together where needed. Record the objective, method and description in the study brief.

Yes. The advanced setup options let you specify which data enters the analysis, such as a particular wave or target group. In the dashboard, editors can further restrict the data base with fixed filters for individual visualisations, sections, pages or the entire dashboard. You can also configure filters for viewers. In the dashboard, they can independently combine several filter characteristics to explore their own view of the shared data without changing the saved analysis.

With an editor seat, you can add a follow-up wave to an existing job on your computer and review the data matching. Upload separate datasets for each wave or a combined dataset containing all waves. The quantitative dashboard results are then updated. For new open-ended responses, unaided mentions and classification questions, you see the required credits; processing starts after your confirmation. Earlier coding is retained. Newly discovered topics can optionally be added to earlier waves after a cost review. Views deliberately set to other waves are preserved.

The interface and tutorial videos are available in German and English. When creating a study, you choose from 18 dashboard output languages; interface and output language can be set separately. In principle, you can use the Insights Agent in any language. Text analysis supports common languages worldwide. Suitability for very small or rare languages depends on the available training data; we review such cases together.

Projects bring your studies together. Search projects and jobs, filter by status and reopen existing dashboards. Frequently used dashboards can be pinned. Archive dashboards and projects to keep the overview clear. Earlier analyses remain accessible for new questions.

Text analysis uses cross-lingual models for common languages worldwide. Responses in different languages can be analysed together, with results presented in one selected output language. Original statements are not translated and remain in their original language. We review very small or rare languages with limited training data separately for your study.

Cauliflower suggests suitable chart types that you can adjust in the settings. Depending on the question and data, these include bars and columns, stacked charts, lines and areas over time, pie and donut charts, histograms, mean profiles, metric cards, ranking, radar and scatter charts, and vertical line profiles. Index charts combine different metrics; dedicated displays support NPS and text analysis, including sentiment charts, driver scatter plots, networks of co-mentioned topics and Summarization Cards. Available drilldowns and additional visualisations in tooltips offer deeper views. Suitable data can also be shown in tables with metrics such as top-two-box, median and mean. Not every chart type is available for every question. The following answers explain brand funnels and index comparisons.

With an editor seat, you can flexibly arrange pages, sections and visualisations. Move elements by drag and drop; duplicate pages, sections or individual visualisations and edit them. In build mode, drag analysed metrics from the library onto the workspace to create new charts and pages. Chart display, colours, layout and density are configurable. Review changes as a draft before applying them; templates and fixed brand colours help maintain a consistent look.

The software recognises repeated questions about several brands as related and automatically combines them in one chart. Index charts compare multiple metrics across brands and countries. Use the comparison tool to compare groups based on existing characteristics such as age or purchase criteria, or combine characteristics into your own target-group variants. Available comparison displays depend on the chart type and data. You can enable significance markers for supported comparisons. Selecting what is displayed does not change the data base; filters and exclusions can change it.

There are three filter types. Editors set fixed filters for a visualisation, section, page or entire dashboard. Temporary filters let viewers combine characteristics for their own view and are reset when the page reloads. Variants are combinations of characteristics saved by editors that viewers can select quickly, such as a particular target group. You can also show or hide brands or statements in a chart. This display selection is distinct from restricting the data base.

Cauliflower recognises ordered scales. For suitable scales, a shortcut combines matrix responses into a top-two-box measure. Top-three-box, bottom-box or your own combinations of values are configurable and can be undone at any time. Scaled metrics can also be converted into dichotomous measures for index charts. NPS modes support 0–10 ratings, assigned response categories and pre-calculated scores from −100 to +100. In the NPS setup, review or choose the appropriate mode and mapping. Depending on the source, dedicated NPS charts show the score, its composition or its development over time; composition cannot be derived from a final score alone.

Yes, you combine and configure metrics per brand or country–brand combination. The average across brands forms the reference with an index value of 100; a heatmap shows deviations from the market average at a glance. Several displays are available, including metric cards, tables, bars, trends and rankings. Drilldowns let you explore brands or metrics in more detail. Changes from a previous wave can also be included directly.

Yes. A brand funnel is built using the index chart and the relevant questions in your study. Combine metrics individually, even when they have different question types and scales: for example scaled awareness, dichotomous consideration, scaled purchase frequency and first choice from a single-choice question. This lets you show several funnel stages together for brands or markets. Different displays also support comparisons across waves. Configure metric selection and transformation to suit the study design.

You can define a weighting column for the analysis. Weighted results are identified in the visualisations and the displayed data base.

For supported group comparisons, you can configure significance markers, confidence levels of 90, 95 or 99 percent, the scope of testing and the reference display. For multiple tests, choose no correction, Bonferroni or Holm–Bonferroni. The procedure depends on the comparison: group markers currently use a t approximation for proportions; suitable unweighted response distributions have a separate pairwise category comparison. Automatic changes from the previous wave are tested separately at a 95-percent confidence level. The explanation at the visualisation shows the test context and reasons for missing markers, such as small or unsuitable bases. A subgroup is currently not tested against a total group that includes that subgroup.

Click the sample size to open the visualisation's data base. It shows the available information on the sample, applied filters, weighting and comparison context. Temporary filters, variants and settings that affect the data are reflected in the displayed context. Differences from the section and your own selection are explained. Small sample sizes or a base that cannot be determined reliably are flagged.

Rename or combine response values directly in a visualisation. For matrix questions, values can be grouped in one step. In Dashboard Studio, you can also centrally recode a column for the corresponding analyses.

Cauliflower offers three analysis types. Unaided mentions such as brands, car models, clothing sizes or spokesperson names are analysed with Unaided Awareness and, where mentions are ordered, Top of Mind. Verbatim Coding structures open-ended responses such as likes, dislikes, associations, NPS reasons or reviews into aspects and aspect categories, adding sentiment and summaries. Classification questions let you ask targeted questions about text and receive analysable answers. The software suggests suitable charts for each analysis type: bars for unaided mentions; hierarchical bars, networks and driver charts for coding; and suitable stacked or unstacked displays for classifications. Classification answers can also be used as filtering and comparison characteristics.

Cauliflower distinguishes positive, neutral and negative statements. You see sentiment for aspects and aspect categories, plus an overall comment sentiment that classifies the entire comment. This lets you examine individual topics and the overall feedback. Filter text results by sentiment; the corresponding original responses remain accessible. Editors can correct a response's sentiment.

Upload your own code frame, have AI create one or build one yourself in the setup's table editor. For AI creation, you can provide additional context and instructions. AI can also extend an uploaded code frame using your data. Aspects and aspect categories can be renamed, split, merged or moved. You can add new aspects or aspect categories after analysis too. Apply changes through reanalysis and restore earlier versions.

Classification questions let you ask targeted questions about open-ended responses, reviews or conversation histories. Options include yes/no, topic assignment as single or multiple choice, scaled responses and extraction of relevant original passages. For example, ask whether a problem was solved, which topics were discussed or how well a case was handled on a scale of 1 to 5. Context and instructions tell the AI what to look for. Answers are linked to evidence and appear in suitable charts. They can also be used as filters and for group comparisons. Processing uses credits per question.

Search original responses and read them alongside their assigned aspects. The classification evidence viewer for conversations requires a personal account and is not available through public sharing links. Saved quotations are highlighted. AI summaries are labelled and refer to an identified data base. Summarization Cards are available per aspect, aspect category or individually filtered selection, such as a target group and selected aspects. Choose bullet points separated by positive, neutral and negative sentiment, extractive summaries containing central clauses in the original wording, or prose summaries based on those extracts.

Driver analysis examines aspects in relation to a target outcome, such as overall sentiment or NPS. Shapley values assess feature contributions in a linear OLS regression model. Partial R² shows the share of variation remaining after accounting for other features that is explained by an additional feature. Kruskal–Wallis tests group differences using ranks. The methods can be configured flexibly. Results can appear as bar, butterfly or scatter charts.

Cauliflower displays co-mentioned aspects as a network, matrix or chord diagram. Adjust how nodes and connections are displayed. This complements the frequencies of individual aspects with their co-occurrence.

Compare aspects and aspect categories using existing group characteristics in a shared index matrix. Tracking studies also have displays of development across waves. These require suitable comparison groups or at least two waves with analysable text results. New text responses are included after the confirmed analysis run. Available comparison displays depend on the chart and data base.

Unaided Awareness analyses spontaneously mentioned brands and groups different mentions appropriately. Classification can also structure other unaided mentions, such as car models, spokespersons or sizes. Top of Mind can be analysed at the same time: it captures the first mention if its order is available in the data. Frequencies and proportions are typically displayed as bar charts. You can choose the reference base and set up several named top-of-mind measures.

When starting text analysis or processing new waves, a cost overview shows the credits required for the selected analyses and classification questions. The run starts after your confirmation. Code-frame reanalysis also uses credits and requires separate confirmation, with a cost notice. Text-analysis credits and the agent budget are different usage measures.

Ask questions such as ‘Which brand has gained since the previous wave?’ or ‘What distinguishes younger from older respondents?’. The agent uses your dashboard's calculated data and content. Where needed, it filters subsamples using combined characteristics, compares groups or waves and calculates measures such as top-two-box. It uses defined analysis tools; its answers are based on the available data and calculated results. Possibilities depend on the data and comparison. Follow-up questions deepen the analysis. With an editor seat, you can also create new classification questions for open-ended responses. A paid classification run starts after a cost overview and your explicit confirmation.

With an editor seat, ask the agent to adjust filters, target-group comparisons, variants or top-two-box settings in existing charts. It carries out clear instructions directly and shows the changes in its response. New charts are initially created as temporary views. Review the result and decide whether to add it to the dashboard. Permanent dashboard changes require an editor seat; registered viewers can ask questions and explore temporary analyses.

Check an answer against the numbers, tables, charts and available evidence used. The agent shows work steps, results and limitations and takes you to the relevant analyses. You can check the data base and filters behind a comparison. For classification questions, original statements and quotation passages help you verify the assignment.

A chat relates to one selected dashboard. Within that dashboard, the agent can use several charts and results to answer a question. You can also reopen older studies and ask new questions there. Combined queries across several studies are in development and are not currently available.

Viewer accounts are free. The agent is included in the plan for active personal viewer and editor accounts. Both use their company's shared agent budget; usage depends on the plan and remaining budget. You can see your consumption and available budget; company admins can also see shared usage. A sharing link alone does not replace a personal agent account. Permanent dashboard changes require an editor seat.

Supported charts, tables and text remain editable in PowerPoint after export. Continue editing charts and tables, adapt text and labels, and change colours and the arrangement of elements. Cauliflower supports almost all visualisations, including text-analysis overviews and driver charts, and carries over defined colours for brands and other entities. Adapt the slides to your story and presentation layout.

With the appropriate management permissions, upload and configure several of your own PowerPoint templates and assign them as company or project defaults. Choose a suitable template for an individual dashboard. Review layouts, colours, logo and placeholder mapping, along with the arrangement of data and comment fields. A preview helps you check the export before downloading it.

For PowerPoint export, select pages, sections and visualisations. You can include title and chapter slides, descriptions, specified filters, data base and number format. Saved variants can be compared on separate slides. The saved dashboard state and selected export options are what count. Temporary filters in your current reader view are not carried over automatically in all cases. Check the data context in the preview and finished file; editors can save target groups needed permanently as a variant or fixed filter.

You can create as many dashboards as you like from one job. This includes a dedicated management dashboard containing only the most important visualisations from a management perspective. You can also save key findings as insights alongside their evidence and arrange them for reports and presentations. Export insights as highlight slides or as a presentation containing only marked content. The order can be set independently of the dashboard structure.

Download table values as CSV or Excel for suitable individual visualisations. Text analysis has dedicated exports: depending on the analysis, CSV or aggregated Excel results, plus response data with codes as Excel or SAV files. A full text download requires authorised personal access to the job and is not available through public sharing links. It contains the available text-analysis data and additional columns. Table exports cover individual visualisations and dedicated text-analysis outputs. A complete tabulation report for the entire study is not currently available.

Download suitable visualisations as PNG; SVG is available for supported chart types, but not for all index, metric or text displays. ‘Copy table’ copies table values to the clipboard. PDF output uses a rasterised print template through the browser's print dialogue; this route is not available for tables. Use Excel or CSV to edit tables further and editable PowerPoint export for supported presentation content.

Dashboards can be accessed through personal logins or sharing links. With the appropriate management permissions, set a password, expiry date, shared pages and export permissions for links. Recipients need no account and can explore the shared views without changing the original. Collections bring several dashboards together under one link. Organise the collection into titled sections and add links to files and further information. Clients and colleagues receive dashboards and supporting content in one place. The Insights Agent and classification evidence viewer require a personal account.

With the appropriate management permissions, revoke a sharing link or replace it with a new one. Manage its password and page selection, and set an expiry date when creating it. This keeps access to shared content under your control.

With the appropriate management permissions, add users to individual projects or remove them. Organise access to the associated jobs and dashboards for different teams and client projects. Sharing-link recipients see the shared content, not your team's members area. Viewer access, editor seats and admin permissions serve different purposes.

Collections let you organise multiple dashboards into titled sections. With the appropriate management permissions, prepare a collection and publish it under one link, optionally with a password and expiry date. Recipients can browse the dashboards provided. Additional link lists point to files and information, such as tutorial videos, documents or contact details. Custom branding depends on the plan and corresponding permissions.

Viewers explore results for free, apply configured filters and download data and visualisations. With a personal account, they can ask the Insights Agent questions and use your company’s agent budget. With an editor seat, you can also create jobs and build and customise dashboards. Separate admin permissions let you manage your team, seats and templates, among other things.

We do not train models on customer data. Our security page provides information on EU hosting and data processing. SSO, private tenant and on-premise are available by individual arrangement. We agree the appropriate scope based on your requirements for authentication, roles and deployment environment.

Guided workflows help you set up filters, comparisons, NPS, significance and insights. Tours explain the interface and how to read results; videos and contextual help support onboarding. You can also arrange a supported 14-day evaluation with us.