What do you like about the product?
Good quality.
In the analysis
- Quality (general)
Surveybot
Better open-ended answers, while the survey is still running.
The Surveybot spots answers that are too brief and automatically asks a fitting follow-up question, right inside your survey tool. Participants describe more precisely what they mean while the interview is still open.
Configurable per questionConnected via API24 languages
Available by individual arrangement.
Dynamic probing
“Good quality” becomes an answer you can use.
Static open questions do not react to what was said. The Surveybot does: if an answer stays too general, it asks a targeted follow-up.
If an answer is already specific, the Surveybot asks no follow-up and the survey simply continues.
What do you like about the product?
Good quality.
SurveybotWhat makes you say the quality is good?
The fabric keeps its shape even after many washes, and the seams hold.
In the analysis
- Quality
- Shape retention
- Workmanship
In Cauliflower you analyse the answer and the follow-up together. “Good quality” turns into concrete aspects such as shape retention and workmanship that you can code, compare and report.
Analyse open-ended answersIllustrative example. The follow-up follows your rules for this question.
How it works
Set up, test, connect.
- 01
Set up
Create a project with your open questions. For each question, define what information a follow-up should draw out and when to ask one.
- 02
Test
Run sample answers through the test area and see whether and how the Surveybot follows up. Adjust the rules before fieldwork starts.
- 03
Connect
Your survey tool calls the Surveybot once per open answer via API. The follow-up comes back and appears as the next question in the questionnaire.
Full control
You decide when and how to follow up.
Rules per question
Define when the Surveybot should always follow up and when never – for example on one-word ratings, or on answers that already give concrete reasons.
Tone and goal
Examples of good and weaker follow-ups set the tone. The goal and study focus of each question decide what the follow-up aims at.
One Surveybot for several concepts
Your survey tool fills placeholders such as {brand} or {product} when it calls the Surveybot. Four concepts therefore do not need four Surveybots.
Fallback follow-up
If the language model does not respond in time, the Surveybot returns the generic follow-up you stored for this question.
24 languages
Maintain questions, rules and examples per language. Missing translations can be generated automatically.
One or several follow-ups
For each answer the Surveybot returns one follow-up. On request it returns several follow-ups at once.
Use cases
Wherever the reasons matter.
NPS and satisfaction tracking
Capture the reasons behind a rating more precisely, wave after wave with the same rules.
Brand tracking and ad recall
Ask what respondents actually associate with a brand or an ad.
Product and concept tests
One Surveybot for every concept in a study, using placeholders for concept or brand.
Ad hoc studies
Also for single studies with only a few open questions.
Technology and integration
One API call per open answer.
- REST API with API key, called server-side from your survey tool
- The response contains the follow-up – or none if the answer is sufficient
- No follow-up for empty answers; optionally none for answers longer than 40 words either
- Fallback follow-up if the language model does not respond in time
We agree the integration together with your scripting team.
Analyse answers and follow-ups in Cauliflower.
Your survey tool records answers and follow-ups together with all other study data. In Cauliflower you code the open-ended responses, including against your own codebook, and report the results in the dashboard. The Surveybot complements your survey software – it replaces neither the survey tool nor panel, recruitment or fieldwork.
FAQ
What project teams want to know upfront.
I ask the same question about several concepts or brands. Do I need several Surveybots?
No. You set up the question once and use a placeholder such as {brand}. When it calls the Surveybot, your survey tool passes on which concept or brand is being asked about, and the follow-up refers to it.
How many follow-ups does the Surveybot ask?
One follow-up per answer by default. On request it returns several follow-ups at once. Whether it follows up at all is controlled by your rules for each question.
What happens if the Surveybot does not respond in time?
If the language model does not respond in time or returns no valid result, the Surveybot returns the fallback follow-up you stored for this question.
When does the Surveybot not follow up?
For empty answers and whenever an answer meets your “never follow up” rule. Optionally it also skips answers longer than 40 words, which experience shows is worthwhile. Specific answers get no follow-up.
Which survey tools work?
The Surveybot is connected via REST API. Your survey tool must be able to call the interface server-side during the survey and show the follow-up as the next question. We work out together how this is done in your tool.
How do I get started with the Surveybot?
With a demo for which we set up a Surveybot with your questions. After that you can start a pilot in a real study. Terms are available on request.
Your next study
Have a study in mind?
We set up a demo Surveybot with your questions and show you how the follow-ups fit your study. After the demo you can start with a pilot; terms are available on request.