What Is Cross Tabulation?
Cross tabulation is a table that shows how the answers to one survey question differ across the groups defined by another. It is how you find out that younger buyers, existing customers or one region think differently from everyone else.
The overall result of a survey is an average of people who do not agree with each other. Cross tabulation, or a crosstab for short, pulls them apart. It works with any closed-ended question.
It answers the question every stakeholder asks second. First: "How many people would buy it?" Then: "Which people?" It is one step in analyzing a market research study.
Key point: the totals tell you what happened. The crosstab tells you who it happened to.
A Cross Tabulation Example
You survey 400 people about a new product and ask how likely they are to buy it. Overall, 175 say likely or very likely, the top two answers on a five-point Likert scale. That is 44%.
Now split that answer by age group.
Likely to buy | 18 to 34 | 35 to 54 | 55 and over | Total |
|---|---|---|---|---|
Yes | 88 | 60 | 27 | 175 |
No | 72 | 90 | 63 | 225 |
Respondents | 160 | 150 | 90 | 400 |
Share likely to buy | 55% | 40% | 30% | 44% |
The overall 44% was hiding a gap of 25 points between the youngest and oldest groups. That changes who you target, where you advertise, and possibly whether the product is worth launching at all.
How to Read a Crosstab
1. Find the groups. They run across the top. Here they are age groups. This variable is sometimes called the banner.
2. Find the question. Its answers run down the side. Here it is purchase intent.
3. Read percentages down each column. 55% of the 18 to 34 group are likely to buy. Compare that with 40% and 30% in the columns beside it.
4. Check the number of respondents in each column. A percentage built on 90 people is much less certain than one built on 400.
5. Compare each column to the total. Groups far from the overall figure are the ones worth acting on.
Use Column Percentages, Not Counts
Raw counts mislead because the groups are different sizes. In the example, 60 people aged 35 to 54 are likely to buy against 27 aged 55 and over. That looks like more than double, but the first group is also much bigger.
Percentages within each group make them comparable: 40% against 30%.
Be clear about which way your percentages run:
Column percentages answer "Of the people in this group, what share gave this answer?" This is what you want for comparing groups.
Row percentages answer "Of the people who gave this answer, what share are in this group?" Here, 50% of likely buyers are aged 18 to 34. Useful for describing your buyers, not for comparing groups.
Mixing them up is the most common error in reading a crosstab.
Is the Difference Real?
Every percentage from a sample has a margin of error, and it grows as the group gets smaller. Splitting 400 responses into three groups makes each one less precise than the total.
At 95% confidence, the three groups look like this:
Group | Respondents | Likely to buy | Margin of error | Range |
|---|---|---|---|---|
18 to 34 | 160 | 55% | ±7.7 | 47% to 63% |
35 to 54 | 150 | 40% | ±7.8 | 32% to 48% |
55 and over | 90 | 30% | ±9.5 | 21% to 39% |
Two comparisons, two different answers:
18 to 34 against 35 to 54: a gap of 15 points. For two groups of this size, a gap needs to be larger than about 11 points to be unlikely to come from chance. This difference is real.
35 to 54 against 55 and over: a gap of 10 points. For these two groups the threshold is about 12 points. This difference could be chance.
Notice that the first two ranges overlap slightly (47% to 63% and 32% to 48%) and the difference is still real. Overlapping ranges do not prove two groups are the same. They mean you need the proper test, which compares the gap to the combined uncertainty of both groups.
The margin of error calculator gives the margin for any group, and the confidence interval calculator gives its range.
The Problem With Small Groups
Each split divides your sample. Split 400 responses by age and you have groups of 90 to 160. Split again by gender and some cells will hold 40 people or fewer.
Under about 30 respondents: treat the percentage as a hint, not a finding.
30 to 100: only large differences mean anything.
Over 100: reliable for most comparisons.
If you already know which groups you want to compare, plan for it before you collect: aim for at least 100 responses in each group that matters. The sample size calculator helps you set the number. How you fill those groups is a samplingdecision: quotas guarantee the numbers, an open link does not.
Clean the data before you split it. A few careless answers weigh more in a small group, so remove respondents who failed an attention check first.
Five Crosstabs Worth Running on Any Survey
1. Customers against non-customers. They almost always differ, and averaging them hides both stories.
2. By age or life stage. The most common source of large gaps in consumer research.
3. By how often people buy the category. Heavy buyers decide your revenue.
4. By awareness or familiarity. People who know your brand well judge it differently from those who have only heard the name.
5. By overall satisfaction. Split any other question by satisfied and dissatisfied customers to see what drives the difference.
Run the crosstabs you planned first. If you run every possible split, some will show a gap by chance alone, and you will have no way to tell which. For questions worth splitting this way, see these survey question examples.
Cross Tabulation and the Chi-Square Test
When a crosstab has more than two rows or columns, the usual test for whether the pattern is real is the chi-square test. It compares the counts you observed with the counts you would expect if the groups did not differ at all.
A small p-value, by convention under 0.05, says the pattern is unlikely to be chance. It does not say the difference is large or important. Always read the percentages alongside the test.
Cross Tabulation in SegmentOS
SegmentOS has a cross-tab and segment analysis tab on the Premium plan. You choose the question that defines your groups, and every other question is broken out by them, with the number of respondents in each group shown. Raw data can be exported as CSV or XLSX on Premium if you want to build tables yourself.
Key Takeaways
Definition: a crosstab shows how answers to one question differ across groups defined by another.
Read percentages down each column, and always check how many respondents each column has.
Smaller groups have wider margins of error. A gap between two groups has to clear both.
Overlapping ranges do not prove two groups are the same. Compare the gap to the combined uncertainty.
Plan your comparisons before you collect, with at least 100 responses in each group that matters.



