Build a balanced MaxDiff design.
Without expensive software.
Type in the items you want to test and get a properly balanced best-worst scaling design you can paste straight into any free survey tool. Everything runs in your browser. Nothing you type here is stored or sent anywhere.
Enter your items, get a balanced design
List the features, messages or value propositions you want to compare, one per line. The tool builds a design where every item appears equally often and every pair of items meets a balanced number of times. That balance is what makes the results comparable.
The balance check
Each cell shows how often a pair of items appears together across your sets. A good design keeps these counts as even as possible, and this matrix lets you check that for yourself.
Run it in any free survey tool
Google Forms, Tally, Microsoft Forms: anything that can show a matrix or single-choice question and export a CSV will do. The nicest way to build a MaxDiff interface in a free tool is the matrix trick below.
- Use one matrix question per set. Make the two rows "Most important" and "Least important", and the columns the items from that set. Respondents see all the items side by side and pick one per row, which is exactly how a proper MaxDiff interface works. In Tally this is the Matrix block, in Google Forms it is the Multiple-choice grid, in Microsoft Forms it is Likert. Keep the default of one answer per row, and mark the question as required. Here is what one set looks like:
What's the most and least important? (1/7)
An outside perspective with no internal politics Findings framed for executive audiences A researcher who understands my industry Confidence the decision is evidence-based Most important Least important One matrix question, built here in a free survey tool: two rows, the set's items as columns. - No matrix question available? Use two single-choice questions per set instead: "Set 1: Which of these matters most to you?" and "Set 1: And which matters least?". The copy button in Step 1 gives you everything pre-formatted for this.
- Keep the sets in order and the wording identical everywhere. Consistency between the survey and your design file is what keeps the data usable. A small heads-up: most free tools cannot stop someone picking the same item as both most and least. Don't worry about it now, those answers just get excluded in analysis.
- Aim for at least 50 respondents from your actual target audience for a simple overall read. Around 80 to 100 gives you comfortably stable rankings for up to 10 or so items, and you'll want more than that if you plan to compare subgroups (say, customers vs prospects).
- Running a bigger study? Generate the design twice (each click produces a fresh one) and split your sample across two survey versions. Everyone seeing the exact same sets in the same order is the main weakness of the free approach, and two versions goes a long way for the cost of one extra survey link. Just keep the exports separate and label which version each came from.
- When fieldwork closes, export the responses as a CSV and keep it together with the design file from Step 1. Those two files are your complete dataset.
The honest small print
This approach is a solid, valid way to rank what your audience cares about, and a big step up from rating scales. But it's fair to know what it doesn't do. Unless you split your sample across versions, every respondent sees the same sets in the same order, so order effects can't average out the way they do in expensive survey software. The results describe your sample as a whole: an average can hide two groups who want opposite things. And the scores are relative, meaning they tell you which items matter more than others, not whether any of them matters in absolute terms. None of this breaks the method for prioritisation. It's just the difference between a scrappy study and a bulletproof one, and worth knowing before you present the results.
The analysis is the tricky part
A well-designed MaxDiff gives you clean data, but the raw answers aren't the answer yet. Scores need to be calculated properly, dodgy respondents excluded, and differences checked for statistical reliability. And then someone has to turn it all into a recommendation you can actually stand behind in front of your team or your board.
That part is my day job. Send me your design file and your raw survey export and I'll work out what your customers are telling you, and what to do about it.
Already collected your data? Analysing an existing MaxDiff is a small, quick piece of work. Starting from scratch? A full study (design, fielding, analysis, recommendation) is a one-week Decision Sprint.
Get your results analysed
A free 30-minute call to look at your question and your data, and figure out the most sensible next step.
Let's chatOr email your files directly: hi@svenjapieritz.com