You can use Compass throughout your workflow. In the editor, it helps you draft and refine surveys; in boards, it supports you with analysis, charts and insights.
Compass doesn’t require in depth knowledge of prompt writing. We’ve designed Compass to produce outputs that align with research best practices whether or not you input a simple prompt or full research brief.
To get the best possible output, it’s important to start with a clear and detailed prompt, the more precise you are, the better the AI can deliver exactly what you need.
General principles for great prompts
Keep these simple rules in mind whenever you’re wiring a prompt in Compass:
Be clear about your goal - say what you’re trying to learn or achieve
Include enough context and examples - mention any background or relevant details
Use plain language - avoid jargon
Be specific - the more precise your request, the better the output
Focus on one task at a time - avoid mixing multiple requests in one prompt
Give examples - they help the AI understand your intent faster
Ask follow-up questions - refine or expand when you need more depth
Review and adjust - if the first result isn’t perfect you can always tweak your prompt slightly and try again!
Not sure, you can also always ask Compass for advice! (e.g. “what do you need to know to create me a really good survey?”)
How to write a good prompt
Creating a survey and questions
1. Be clear about your research goal
Start by describing what you want to learn, or any hypotheses you have. What is the purpose of your survey?
For example:
“I want to understand customer satisfaction with our new product feature.”
“I need to identify what drives brand loyalty in our market.”
This helps Compass focus the design, tone and structure of your research around your real objectives.
2. Describe your target audience
Tell Compass who you want to hear from. Include demographic and behavioural details where possible, such as:
Age range
Gender
Location
Purchase habits or behaviours
For example:
“UK adults aged 25–45 who have bought a smartphone in the last six months.”
“NatRep sample in the UK”
A well-defined audience helps ensure the questions are relevant and the data is meaningful.
3. List your key topics
Highlight the themes or focus areas your research should explore.
Examples include:
“I’d like the survey to focus on product usage, satisfaction and likelihood to recommend.”
“Include questions about brand awareness, purchase frequency and price perception.”
“Cover topics like customer loyalty, reasons for churn and satisfaction with our service.”
This ensures the AI builds questions that align with the insights you need.
4. Mention question types or structure
If you prefer certain question types or a specific flow, include that too.
Examples:
“Include a mix of multiple-choice and open-ended text questions.”
“Start with a screening question, then ask about satisfaction and intent to recommend.”
Specifying this helps tailor the survey experience to your preferred structure.
5. Add background or context
Include any details that explain the background of your project, company or product.
Examples:
“We’re a subscription meal kit service exploring reasons for customer churn.”
“We’ve recently rebranded and want to test awareness among existing users.”
Context helps the AI design relevant and unbiased questions.
Example of a good prompt:
“I want to understand why, how, where people in the UK and US aged 25-40 are consuming low/no alcohol beer and what the triggers and barriers are to consumption, as we are looking for opportunities to launch into this category. Start with a screening question on users who have purchased alcohol in the last 6 months.”
Creating charts and analysing your results
1. Clear questions or describe what you want to visualise
Explain which data or question you want to turn into a chart, avoid generic questions like “what does the data say?”
Examples:
“Show how satisfaction scores vary by age group.”
“Which age group is most likely to buy low- or no-alcohol drinks?”
This helps Compass identify the right data points and choose the most relevant way to display them.
2. The type of chart you want
Mention a preferred chart format if you have one.
Examples:
“Use a bar chart to compare results.”
“Create a pie chart showing percentage breakdowns.”
3. Any filters or comparisons
Add context about how you want to filter or compare the data.
Examples:
“Compare results between new and returning customers.”
“Compare 18-30 & 30+”
“Show only responses from participants in the UK.”
This helps Compass focus on the most relevant segments to you and highlight meaningful differences.
4. Your goals or reasons for asking
The more context you can give will help Compass to understand your intent and tailor the responses and charts to answer your real question.
Examples:
“I want to understand whether younger shoppers are driving the growth of the low/no alcohol category.”
Not quite sure what to ask yet?
You can ask Compass for inspiration when you are unsure how to begin your analysis. It can help you spot themes, suggest angles or highlight patterns you may want to explore.
Examples:
“I am not sure where to start. Can you give me a few ideas for what I could explore in this dataset?”
“Which questions would be interesting for cross-analysis”
How we’ve trained Compass
Compass has been trained specifically for consumer research by researchers. Here’s what you can expect when you use it:
Compass has been tailored for research, so always takes in to account research best practices.
It has structured responses, so it will usually reply in a clear organised format. This means any insights are easy to read and it will always give you helpful next steps.
Compass has a friendly and supportive tone, it is trained to sound like a research teammate that is approachable, curious and helpful. It always uses everyday words and avoids jargon in UK english.