Pietrees and matrices as answers land.[1]
Run the tree test you came for and pay only for the methods you run. When the next project needs interviews or a diary, they go in the same study, on the same participants.
30 study-days free. No card.
Tree testing, card sorting (open, closed and hybrid), first-click, five-second and preference tests, and surveys with SUS, NPS and UMUX-Lite.
Tree testing, scored down to the pietree.
Tree testing checks whether people can find things in a site's structure. They see the navigation as plain text, so the visual design can't help them.
Here's a three-question tree test on Stroop's own site. Your path is scored the way a study scores it and drawn into the pietree beside 40 sample participants.
Question 1 of 3
Where would you check what a diary study costs?
Answer a question and your path is drawn into the pietree beside it, with 40 sample participants.
Card sorting, down to the dendrogram.
Card sorting asks participants to group your content into categories, their own (open), yours (closed) or both (hybrid).
The similarity matrix is ordered by the clustering, so groups read as dark blocks on the diagonal. The dendrogram uses average linkage, the same as R's hclust. How the analysis works
Answered on the phone in their hand.
A tree test, a card sort or a survey opens from a link with nothing to install and no account. Every screen carries your logo and colour, and the time estimate is on the first one.
On a laptop the same study runs in the browser, and card sorts get drag and drop with the whole board in view.
A first-click test and three questions, after the tree.
Quick methods stack in one study: the tree test, a first-click test on the homepage design and a few follow-ups, answered in one sitting by the same people.
Stroop is the all-in-one research and testing platform. Run usability tests, prototypes, surveys, tree tests, diaries and IATs in one study, and get results back transcribed, scored and charted, with every finding linked to the session it came from.
First clicks, recall and preference, charted on arrival.
Every chart has a table view and an interval, and every open answer stays one click from the person who wrote it.
First-click testing on any design.
First-click testing measures where people click first when given a task on a design. A right first click is a strong sign the task will be finished.
Five-second tests, counted word by word.
A five-second test shows a design for five seconds, then asks what people remember, to check what a page says at a glance.
Preference tests with the reasons kept.
A preference test shows two or more designs and asks which people would rather use, and why. Position is randomised and checked for bias.
Surveys with SUS and NPS scored.
A survey here is a questionnaire with branching logic and standard instruments, SUS, NPS and UMUX-Lite among them, that score themselves as answers arrive.
Where it stops.
A week-long tree test with any number of responses is $21. See every rate
- A tree test that fills in an afternoon still runs the 7-day minimum.
- Word counts on five-second and open answers group words by stem. They don't read meaning, so cheap and good value stay apart until you tag them.
- A card sort holds up to 200 cards. Past about 60, participants tire and the matrix gets noisy, so split the content into two sorts.
Questions
Should I run an open or a closed card sort?
- Open when you don't know the categories yet: participants name their own groups, and the similarity matrix shows what belongs together. Closed when you have categories and want to see whether content fits them. Hybrid gives people yours and lets them add their own, which helps when you suspect one is missing.
How many participants does a tree test need?
- About 50 per task gives a success rate with an interval of roughly plus or minus 13 points, enough to compare two trees. Thirty will show you where people go wrong. Stroop prints the adjusted Wald interval beside every rate, so a task that needs more people looks like it.
What does directness mean in a tree test?
- Directness is the share of participants who reached an answer without going back up the tree, whether the answer was right or wrong. High success with low directness means people found it after hunting around, so the first label they tried wasn't the one that held the answer.
How are SUS and NPS scored in a survey?
- SUS follows Brooke (1996): odd items score the answer minus one, even items five minus the answer, and the sum is multiplied by 2.5. NPS is the share of 9s and 10s minus the share of 0 to 6, printed with its margin of error. Both score as responses arrive. The SUS note has a worked example.
Where do the participants come from?
- You invite them, from your own customers, a client's list or a panel such as Prolific, and pay them directly. If you want a panel on the same invoice as the tree test, Optimal Workshop and Lyssna both sell one.