Questionnaire

You build your questionnaire on the “Study” tab. Below are the key features and tools available to you.

Screener questions

Use screeners to limit your sample to your target audience. A screener comes at the very start of the study and filters out twins who don’t match. Mark each response option as Pass or Exclude. You can add up to 2 screener questions.

Note

Screeners are only available with a fresh, new sample (not saved samples).

Best practices for screeners

1. Avoid yes/no questions. A binary question leaves room for interpretation and cuts out people who only partly match. Ask about frequency or recency within a clear time frame instead, and exclude only the options you really need to.

Avoid

Do you drink coffee? Yes / No

Better

In the last 3 months, how often have you bought coffee beans or ground coffee?
Never (Exclude) · Once (Pass) · 2–3 times (Pass) · Monthly or more often (Pass)

2. Use one criterion per question. A question that combines conditions — “Do you own a car and plan to replace it this year?” — screens out anyone who meets only one of them. If you need both, use your two screeners for them and choose how to combine them (AND / OR).

3. Move situations into the study context. If a condition describes a situation rather than who the person is — doing the groceries, scrolling through social media on their phone — you usually don’t need a screener for it. Put it in the study context instead, for example: “You are doing your grocery shopping and browsing for household cleaning products.” That keeps both screeners free for the criteria that really define your audience.

4. To exclude buyers or users, ask about recency.

ExampleWhen did you last buy a product from [Brand]?
Within the last 6 months (Exclude) · 6–12 months ago (Pass) · More than 12 months ago (Pass) · I have never bought from [Brand] (Pass)

When only a few twins pass

Expect low incidence for niche audiences. If only a few twins pass, it usually means the audience is small in the real population: Elaiia only creates twins whose characteristics coexist in the real world, never fabricated profiles — that is what keeps the simulation grounded in reality. If the incidence is low, increase your sample size, or run the study again with a new sample and look at the results in aggregate.

Instructions

Instruction blocks allow you to provide twins with additional context, define a situation, or guide how they should approach the questions that follow. Instructions are shown to the twins but are not answered or displayed as questions in the results.

How instructions affect the twins

When you create a new twin sample, the instructions become part of the core context used to create the twins.

When you reuse saved twins, newly added instructions work more like directions that guide how the twins should answer the questions in the new study.

Instructions apply to the questions below them

The position of an instruction block matters. Its context applies to the questions placed after it in the study.

For example, an instruction placed before a concept evaluation can establish the situation in which the twin encounters the concept:

Example instructionImagine you are scrolling through social media on your phone during your commute home. You see an advertisement for a new banking app that automatically categorises your spending and suggests ways to save money.

The questions below this instruction will be answered within that context.

If an instruction block is placed at the very end of a study, there are no following questions for it to contextualise. When creating new twins, the instruction may still influence how the twins are formed, but it will not provide context for any answers.

Limiting instructions to specific questions

By default, twins continue to consider previously provided context when answering later questions. When an instruction should apply only to a particular part of the study, clearly define its scope within the instruction itself.

Example instruction with scopeImagine you are scrolling through social media on your phone during your commute home and see the following advertisement. Apply this context only to the next two questions. After those questions, return to your general perspective and do not use this scenario when answering the remaining questions.

This is especially important when a study contains multiple scenarios, topics, or evaluation tasks. Explicitly stating where an instruction starts and ends helps prevent the context from influencing questions where it is no longer relevant.

Tip

Instructions and Scenarios & Context are closely related — plan your scenarios in advance, then implement them as instruction blocks here.

Question types

Elaiia supports a range of question types. Choose the type that best fits what you’re trying to measure.

Likert Scale

Measure agreement, satisfaction, or intensity on a defined scale (e.g. 1–5 or 1–7).

Example How satisfied are you? 1 = Very dissatisfied … 5 = Very satisfied

Metric

Collect a single numeric value — prices, quantities, frequencies.

Example How much would you pay? Answer with a single number in CHF.

NPS

Net Promoter Score — the standard 0–10 recommendation likelihood question.

Example How likely to recommend? 0 = Not at all … 10 = Extremely likely

Choice Frequency

How often does the respondent choose, use, or encounter something?

Example How often do you purchase chips? Never / Monthly / Weekly / Daily

A/B Test

Present options side by side and ask twins to pick a winner.

Example Which packaging design do you prefer? Option X / Option W

Idea Generation

Open-ended creative prompt — twins brainstorm or generate new ideas.

Example What new flavour would you create? Describe in 2–3 sentences.

Concept Evaluation

Show a concept and collect structured feedback — appeal, intent, likes and dislikes.

Example How appealing is this concept? What would you change?

Rank Order

Ask twins to rank items by preference, importance, or priority.

Example Rank from most to least important: Price, Quality, Brand, Design

Usage Exploration

Explore how, when, and why people use a product or service.

Example Describe a typical situation in which you use this product.

Matrix

Rate multiple items on the same scale in a single grid. See Matrix Questions for how to write one.

Example Rate each brand on: Quality, Value, Trust (1–5 per row)

Randomising questions

To randomise questions, create question groups. Right-click on a question or use the context menu.

1. Select questions
2. Create Group
3. Toggle “Randomise”

Image descriptions

When adding images, always add an image description. This text is what the AI twins “see” — without it, the image may not be interpreted correctly.

AI validation

Highly recommended before every study launch

Elaiia’s built-in AI validation reviews each question for specificity, clarity, and answerability. It catches ambiguity, vague phrasing, and structural issues that reduce data quality — and suggests concrete improvements. This step takes seconds and can significantly improve the quality of your results.

Select “Validate questions with AI” after creating all your questions. You can compare the original wording with the suggested improvement and choose which version to keep.

Tip

You don’t have to adopt every suggestion — review each one and decide what best fits your research objectives.