Build any questionnaire on a map you can see: every question, route and screen-out drawn as a line, so you can check the logic before anyone takes it. Bring in a questionnaire you already have from Word or PDF. Eleven automatic quality checks then judge every response, and a rejected panel response is not charged.
Add questions to the map, or upload a Word or PDF questionnaire and Graham lays it out for you, routing, screen-outs and quotas included, as a new draft you check.
Take it yourself on a test link, preview any question, and send hundreds of simulated respondents through the logic to see where they go and how long it takes.
Share your own link or order a panel. Results chart every question, with cross-tabs, flows, drop-off, and Graham's report.
Most survey tools hide skip logic in a settings panel on each question, so the only way to check it is to take the survey again and again. Here the whole questionnaire is one map. The screener sits in its own box above the study, and every answer that routes somewhere has its own line.
Every response is judged when the respondent finishes, on times taken on our server rather than their browser. The answers a check looks for, like the right reply to an attention question, never reach the respondent's screen. A weak signal never rejects anyone on its own: it holds them for you to look at.
Speeding, timed against how much each screen asks them to read. Straight lines down a grid, zig-zags, and always picking the same position.
Attention checks, made-up answer options, ticking every box on a screener, two answers that contradict each other, and numbers nobody could believe.
Keyboard mash and one-word replies in open text, text pasted or copied from another respondent, and someone who completed, or was screened out, coming back to try again.
Each check either rejects on its own, holds the response for you to review, or is simply noted for you to judge; two holds together reject. In a simulation of 2,000 careful people and five kinds of bot, about 91% of the careful people passed and every simple bot was rejected. More in the guide: How to spot bad survey data.
Each question gets the chart that suits it, with its base beside every percentage. Cross any question by another's answers with significance marked, follow how answers flow from one question to the next, and see exactly where people stopped.
Graham builds the first draft from your questionnaire, and when the responses are in he writes the report: the headline, the differences between groups that matter, and the themes in what people wrote in their own words.
Share a link with your own customers, or order respondents from our panel partner. Standard demographics on the map become the panel's targeting, so people who do not fit are never sent and never billed. The audience page shows what the panel will be asked for, and anything that has to be screened inside the survey instead.
See pricingUse it for concept tests, usage and attitude studies, brand tracking, customer satisfaction, or any questionnaire with logic. If you need to rank a long list, Best–Worst or pairwise will give a sharper answer than a rating grid. If you are still finding out how customers group a list, start with a card sort.
Sixteen: single choice, multiple choice, dropdown, image choice, Likert agreement, star or smiley rating, NPS, slider, matrix grid, short text (with email, phone, ZIP, number or pattern checks), long text, number, date, rank order, consent and recontact, and text or media screens with pictures or video.
Yes. Upload a Word document or a PDF, or paste the text. Graham reads it and builds a new draft survey on the map, with its routing, screen-outs and quotas, and lists anything he could not map so you can finish it by hand. Nothing you already have is overwritten.
Every answer that sends people somewhere different has its own line on the map. If and then cards, random splits, quotas, blocks and loops are cards you place. The builder flags a question nobody can reach, or a route with nowhere to go, while you build.
Eleven checks run on every finished response: speeding, grid straightlining, attention checks, made-up options, screener over-selection, contradicting answers, junk open text, copied text, answer patterns, impossible numbers and duplicate people. A panel response that is rejected is sent back and you are not charged for it.
Yes. Take it on a test link as many times as you like, preview any question, and simulate hundreds of respondents to see where the logic sends them and how long the survey takes. Test answers are kept apart from live ones and can be cleared.
A survey is a study like any other on your plan. The free plan includes one trial study with up to 10 participants, and paid plans start at $49 a month with unlimited studies. Panel respondents are priced per response on top.
Compare survey software · More in the guides: How to spot bad survey data · Which study do I need?
Start from a blank map or your own questionnaire. Graham will help you get going.
Start your free study