> ## Documentation Index > Fetch the complete documentation index at: https://support.deepfield-ai.com/llms.txt > Use this file to discover all available pages before exploring further. # Configure and compare concept tests > Choose concepts, rotation and exposure, test each route, and compare results using the right bases. Use a concept loop to ask the same questions about several ideas, products or versions. Decide whether each participant sees one concept or several before building the questionnaire. ## Add concepts to a section **1. Define the concepts** Open the section's settings and select **Add concepts**. Name the loop and each concept distinctly. Attach the stimulus each concept needs. A variant group presents one version of its idea to each respondent. **2. Choose who sees what** Set **Concepts per respondent** and the rotation. **Balanced** balances exposure and position; **Random** draws a random order; **Fixed order** keeps the authored order. Balance within an earlier single-choice answer or categorical variable when the design requires it. **3. Set exposure** Choose whether the concept appears on every question, on the first then through a button, through a button only, on click, or as a timed first impression. Timed exposure requires media. Check video playback requirements separately. **4. Add the repeated questions** Write the questions in the looping section. Each repeats for the concepts assigned to that respondent. Use a question's concept-visibility setting when its stimulus needs different treatment. A loop can take its items from an earlier answer instead of a fixed concept list. Nested loops repeat the questions for combinations of the outer and inner items. Name every loop so its references remain understandable in the questionnaire, Responses and charts. ### Show the concept at once or on a click A question's **Concept on this question** setting offers, among others, **Show**, **As a button** and **On a click**: * **Show**: the concept is visible above the question as soon as the page opens. The respondent has nothing to do. * **As a button**: a **View again** button under the question opens the concept in a window. Opening it is optional: the respondent can answer and continue without it. * **On a click**: the concept stays hidden behind a **Show** button above the question. The respondent cannot continue until they select it, even after answering. To let respondents look at the concept again only if they want, choose **As a button**. On a heatmap set to **On a click**, the image or text to mark appears only after the click, so marking time is counted from the click rather than from the page opening. ## Compare concepts directly Use the side-by-side concept display on a matrix when the participant should rate concepts together. Use an A vs B task for head-to-head choices. Picture cards change how choices are presented; they do not turn an ordinary selection into a concept loop. A later question can take a loop's concepts as its options. Check that its source actually supplies the items the comparison needs, especially after filters and conditional assignment. ## Test the design Save and test the assigned concepts, their order, conditional images, exposure and repeated questions. Exercise each cell of a monadic design, in which a respondent sees one concept, and more than one route through a sequential design. Test Back and translated concept names. Inspect the recorded concept order and the answers in Responses. Balanced rotation aims to distribute exposure; it does not guarantee identical completed sample sizes. Screen-outs, conditions and incomplete responses can change the achieved bases. In a monadic design with quotas, balance the rotation within the quota questions so that each profile is split evenly across the concepts. See [Set quotas for a monadic or split test](https://support.deepfield-ai.com/recruiting/quota-management.md#set-quotas-for-a-monadic-or-split-test). ## Read the comparisons In Charts, a repeated question is read by concept. Check the base for each concept, the filters and weighting before comparing it. A monadic assignment variable that simply repeats the concept columns is omitted from the plan's comparison table. When one respondent contributes several concept answers to a table axis, the table distinguishes answer counts from respondent counts. These repeated answers are not independent people: significance tests and confidence intervals are not computed for that repeated-record table. See [Read statistical significance](https://support.deepfield-ai.com/analysis/reading-significance.md). ![Review each concept, the rotation and the questions repeated for it.](https://mintcdn.com/deepfieldai/dXwxAZySujI3sywX/images/building-your-study/concept-testing.png?fit=max\&auto=format\&n=dXwxAZySujI3sywX\&q=85\&s=67a99e4b7eeb0e02f89053140cc1ef73) Review each concept, the rotation and the questions repeated for it. ## Related * [Add stimuli and concepts](https://support.deepfield-ai.com/building-your-study/stimulus-materials.md) * [Configure conditions and routing](https://support.deepfield-ai.com/building-your-study/advanced-logic.md) * [Build an analysis plan](https://support.deepfield-ai.com/analysis/analysis-plan.md) * [Build a crosstab](https://support.deepfield-ai.com/analysis/crosstabs.md) This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.