Guides
The questions people actually ask before running a study: how many participants they need, which method answers which question, and how to read what comes back. Written for the people who have to make the decision, not for a methods exam.
How many participants do you need for a card sort?
About 15 people gets you a result that holds. 20 to 30 makes it firm. Beyond 30 the picture stops changing, and here is the evidence for each of those numbers.
Card sorting →Card sorting vs tree testing: which one, and when
Card sorting designs a structure. Tree testing proves one. They answer different questions, they run in a particular order, and using the wrong one wastes a round of research.
Findability →Open, closed and hybrid card sorts explained
Open sorts discover a structure, closed sorts validate one, hybrid sorts do a bit of both. Picking the wrong one answers a question you were not asking.
Card sorting →Writing card sort items people can actually sort
Most disappointing card sorts were decided before anybody sorted anything. The items were internal language, or duplicates, or a mix of things at different altitudes.
Card sorting →Reading a dendrogram without a statistics degree
The tree diagram a card sort produces looks like a phylogeny and frightens people off. It is simpler than it looks, and reading it left to right is the whole trick.
Card sorting →Why rating scales fail on long lists
Ask twenty questions on a 1 to 5 scale and almost everything comes back important. The problem is not your respondents, it is the instrument.
Best-Worst (MaxDiff) →MaxDiff vs conjoint: what each one is actually for
Both force trade-offs, both produce numbers that look similar, and they answer different questions. One ranks what matters, the other predicts what people will buy.
Best-Worst (MaxDiff) →Finding unmet needs with an opportunity matrix
Importance on one axis, satisfaction on the other, and the gap between them is the opportunity. Ranking by importance alone sends teams to build what customers are already happy with.
Pairwise →What a tree test tells you that analytics cannot
Analytics shows you what people did on the site you have. A tree test shows you what they were trying to do, where they looked first, and which label sent them the wrong way.
Findability →How to spot bad survey data before you pay for it
Speeders, straight-liners and grab-and-dumpers are a fact of paid panel research. What separates a usable study from a wasted one is catching them automatically, and not being billed for them.
Pricing →
Not sure which study you need?
Answer one question about the decision you are trying to make, and we will point you at the method that answers it.
Find the right study