Analysis
Read your first crosstab
Cut any question by any segment with a crosstab, read counts and column percentages, and use significance testing to know which differences are real.
A crosstab answers the question a single-question chart can't: does this result differ by segment? Instead of "62% chose Option A," a crosstab shows you 62% overall but 71% among women and 54% among men — the cut that turns a number into a finding. Here's how to build and read your first one in Surveti.
Screencast coming soon — the steps below cover everything you need now.
What a crosstab is
A crosstab puts one question in the rows and another question in the banner columns, then shows how the row answers split across those columns. The rows are what you're measuring; the banner is how you're slicing it.
Rows vs. banner
A good habit: put your outcome question in the rows (the thing you care about) and your audience question in the banner (age, region, plan, segment). Read percentages down each column to compare segments.
Build the crosstab
Open reporting for your survey
From a survey with responses, go to Reporting. Every question already summarizes automatically — no export required.
Choose your row question
Pick the question you want to analyze — say, "Which plan are you considering?" — as the rows of the table.
Add a banner question
Add a second question as the banner columns — for example a segment like company size or region. Surveti cuts the rows by that question.
Read counts and column percentages
Each cell shows a count and a column percentage. Column percentages add to 100% within each banner column, so you're comparing like with like across segments.
Read it with confidence
A gap on screen isn't automatically a real gap — small samples wobble. On Professional and above, add significance testing (chi-square and z-tests) with market-research-style column letters, so a difference that clears the bar is marked and one that doesn't isn't.
A crosstab with column percentages and significance letters marking where segments differ.
Watch your base sizes
A striking percentage on a tiny column can be one or two people. Check the count in each cell before you quote a result — significance testing helps, but a low base is a low base.
Take it further
Need more than a simple cut? On Research Team, banners can be weighted and nested, so you can read, say, region within age group and correct for sample skew at the same time. And when you want to hand the analysis to someone else, build it into a shareable report or export the underlying data to CSV, XLSX, or SPSS.