Key concepts
- Item: one thing being scored, with a text label and an optional image. A question can have a single item, such as one “How likely are you to recommend us?” question, or many.
- Point scale: a fixed set of labelled scale points. You start from a preset, such as Agreement or NPS, or write your own. Each point has a number (its value) and a label.
- Slider scale: a numeric range with a Minimum, Maximum and Step, and a label under each end. The slider is a scale type of the Carousel (Grid) question, not a separate question type.
- Scale direction: which end of a point scale is the best answer. It decides how results are read, not what participants see.
- Top-2-box: the share of answers on the two most favourable points of a point scale.
Setting up a Carousel (Grid) question
- Add a Carousel (Grid) question to the topic guide and write the question, such as “How much do you agree with each statement?”
- Choose a scale from the scale menu next to the question type: a preset, Slider · Custom range, or Write my own. See Point scales and Slider scale.
- Under Items, use Add item to add the items participants score. Rename, delete and drag items to reorder them. Use Add image on an item to show a picture with it.
- Choose Fixed order or Random order for the items. Random order shuffles the items for each participant. Pin an item to keep it in its position while the others shuffle.
- Optionally attach a stimulus to the question, such as a concept the items are scored against.
- Review the probing settings (see Probing).
- Test the interview on a phone and on a computer.
Translations
When a study’s interview languages differ from its reporting language, use Translations next to Items to review each item’s label in every language. Preset point labels use built-in translations. Custom point labels and slider end labels are translated automatically.Showing an item to some participants
An item can have a routing rule, so it is shown only to participants whose earlier answers match the rule. Items without a rule are shown to everyone. See question routing.Point scales
Presets
The point values run from 1 upwards, except NPS, which runs from 0 to 10. To change a preset’s number of points, pick a different preset; points can only be added or removed one at a time on a custom scale.
Custom scales
Choose Write my own to start a custom scale with 5 blank points. Use Customize labels to fill in each label, and add or remove points to get between 2 and 11 points. Points are numbered from 1 in order. Every point needs a label before you can launch. Two points with the same label trigger a warning, because participants would see identical choices. You can also use Customize labels on a preset. Changing any of its labels turns it into a custom scale. On the NPS preset, that means the question no longer gets an NPS score (see NPS).Scale direction
The Best control next to Scale sets which answer is the best one: Last answer (the highest value, the default) or First answer (the lowest value). Use First answer for a scale written top-down, such as 1 = “Very relevant to me”. Direction changes how results are read: which item is best and worst, which end is the top-2-box, and the chart colours. It never changes the scale participants see, so you can change it at any time, including after fieldwork starts. The NPS preset always reads 10 as the best answer, and its direction can’t be changed.What participants see
- The question is shown above a progress marker, then the current item: its image, if any, its label, and the scale.
- The scale points are buttons in one row when the screen has room, and a stacked list on narrower screens. The Stars preset is shown as stars. On a numbered scale with worded ends, such as NPS, the end labels are printed under the row.
- Tapping a point records the answer and moves straight to the next item. There is no separate confirm step.
Analyzing point-scale results
Open the question in Question Analysis. The Rating analysis card shows how answers spread across the scale, with one stacked bar per item. Each item is calculated on its own answers. If the question is asked once per image of a randomized stimulus group, the card shows one chart per image, so scores for different concepts are never merged. The card can also show these caveats:- Low sample: fewer than 30 people answered an item. Read it as a signal to explore, not a result. In a diary study, this counts people, not entries.
- Mixed scale: scale points were added after answers started coming in, so figures aren’t comparable across the point where the scale changed.
- Different numbers of answers per item: for example, an item added during fieldwork, or a participant who stopped partway through. Each bar is calculated on its own item’s answers.
Top-2-box and bottom-2-box
Top-2-box is the share of an item’s answers on the two most favourable points; bottom-2-box is the share on the two least favourable points. Which end is favourable comes from the scale direction. On a “First answer is best” scale, the top-2-box is the two lowest values.NPS
An NPS score is calculated only for a question on the NPS preset. Answers of 9–10 are promoters, 7–8 passives and 0–6 detractors. The score is the percentage of promoters minus the percentage of detractors, from −100 to 100. In a question with several items on the NPS preset, the score pools the answers of every item. No NPS is calculated on any other scale, including a custom 0–10 scale or a relabelled NPS preset.Talk to Your Data
Talk to Your Data can answer questions about point-scale results. It reports each item’s number of answers, mean and top-2-box, the best- and worst-scored items, the low-sample and mixed-scale caveats and, on the NPS preset, the NPS with its promoter, passive and detractor split. It can draw a stacked bar chart, or a heatmap of items by scale point for a question with two or more items. For a question asked once per image, ask about one concept at a time. Each item of a point-scale question is also a facet, so you can filter by it or compare it against another facet, such as a rating by gender.Slider scale
A slider lets participants answer each item by moving a handle along a numeric range, such as 0 to 100, instead of picking a worded point. Use it when you want a fine-grained or continuous score per item.Adding a slider
- In the question type menu, go to Scale type and choose Slider · Custom range. You can also switch an existing Carousel (Grid) question’s scale to Slider from its scale menu. A new slider starts at 0 to 10 in steps of 1.
- Set the Minimum, Maximum and Step.
- Fill in Label for the minimum and Label for the maximum. Each label describes what that end means, not a midpoint.
Range rules
- The minimum must be lower than the maximum, and the step must be greater than zero.
- The step must divide the range evenly, so the maximum is a position on the slider. For example, 0 to 100 in steps of 5 works; 0 to 100 in steps of 3 does not.
- A slider can have at most 1,001 positions. Use a larger step if the range and step give more.
- Values can have up to four decimal places.
- Both end labels are required before you can launch.
Word every item in the same direction
A slider has no direction setting. The end labels are the only thing that says which end is the good one. Word every item so that the maximum end means the same kind of answer for all of them. Mixing directions in one question makes the item means impossible to compare.What participants see on a slider
Participants see one item per screen, as on any Carousel (Grid) question.- The handle starts in the middle of the range and is greyed out. The middle is never recorded as an answer on its own.
- Tapping anywhere on the track moves the handle there. Participants can also drag the handle, or use the arrow keys on a keyboard.
- The selected value is shown as a large number above the track, with the end labels under each end.
- Confirm stays disabled until the participant touches the slider. Confirming records the value and moves to the next item.
- The track shows a tick mark for every position when the slider has 21 positions or fewer (for example 0 to 100 in steps of 5). Finer scales show a plain track, but the handle still snaps to every step.
Analyzing slider results
Open the question in Question Analysis. Answers are kept as the exact numbers participants chose.- Single-item question: the Mean and Median, the 95% interval of the mean, and a histogram of the answers.
- Multi-item question: a chart comparing every item’s mean on the shared scale with its 95% interval, a table with each item’s Mean, Median and 95% interval, and a histogram per item.
- The 95% interval describes uncertainty about the mean. It is not the range most participants answered in.
- The histogram has one bar per position when the scale has 31 positions or fewer. Finer scales are grouped into 10 equal ranges.
- Items answered by different numbers of people (for example, an item added during fieldwork) are each calculated on their own answers. The card says so when this happens.
- When fewer than 30 people answered, the card warns that the average is not reliable. Read it as a signal to explore, not a result.
- Which end is favourable comes from your end labels. On a scale from “Excellent” to “Poor”, a higher mean is a worse result.
Choosing a point scale or a slider
Use a slider when the question asks for a number: a likelihood out of 100, “how much out of 10”, or any score where fine distinctions matter. Use a point scale when the answer reads naturally as words, such as agreement, satisfaction or frequency. A point scale stays the default for these. Use a point scale, not a slider, when you need these:- NPS: use the NPS preset. A 0–10 slider has no NPS, and its answers are never grouped into promoters, passives and detractors.
- Top-2-box: slider answers are never converted to scale points, so there is no top-2-box.
- Screeners, randomized stimuli or facets: see Limitations.
Probing
Open Probing Config on the question to review its probing. Probing runs after the participant has scored every item, in two phases:- Item follow-ups: the answers are split into three groups: the lowest point, the middle, and the highest point. On a slider, the groups are the lowest tenth of the range, the middle, and the highest tenth. The interviewer asks one follow-up about a randomly chosen item from each group, up to three in total.
- Closing probe: an optional final question about the set as a whole. It runs only when you give the AI a goal, such as “Ask which statement they found hardest to judge and why”. The follow-ups setting controls how deep it goes. On a slider, setting follow-ups to none turns off probing entirely.
Exports
In a custom export, choose Rating (per item) to add one column per item. Each column holds that participant’s answer: the value of the chosen point on a point scale, or the exact number on a slider. A question asked once per image of a randomized stimulus group gets no rating columns, and the export settings say so.Editing after participants answer
Once a participant has answered in a real (non-test) interview:- You can’t replace the scale: no switching to another preset, to a custom scale, or between a point scale and a slider. On a slider, you can’t change the range.
- You can’t delete an item or a scale point that has been answered.
- You can still edit labels, add items, add points to a custom scale, reorder and pin items, and change the scale direction. Adding points after answers start makes the analysis show a mixed-scale caveat.
Limitations
- Screeners: point scales only. A Carousel (Grid) question in a screener uses a point scale. Set Qualifying answers (a From and To point) on each item that should qualify participants. Items left on Any response don’t affect qualification, and at least one item needs a range before you can launch. An answer outside an item’s range screens the participant out. A screener question can’t use the slider scale.
- Routing on answers: screeners only. A routing rule can be based on a Carousel (Grid) answer only when the question is in the screener. You can still route to or past a Carousel (Grid) question anywhere. See question routing.
- Slider: not with randomized stimuli. A slider can’t be used on a question, or in a section, that has randomized stimuli attached. That question is asked once per stimulus, and slider results can’t yet tell those answers apart, so scores for different concepts would be merged. Use a point scale (a 0–10 or 1–7 custom scale keeps the detail), or ask the slider once in a section without randomized stimuli.
- Slider: no NPS, top-2-box or facets. Slider questions don’t create facets for filtering or segmenting. Their analysis reads the answers directly.
- NPS only on the NPS preset. Custom and relabelled scales never get an NPS score.