You have survey results on your screen and two questions. First: how many people gave each answer? Second: does that pattern differ between groups, such as new and existing customers? The first question calls for one table and the second for another, and choosing the wrong one is a common reason a summary feels thin or cluttered.
This article works through both on the same data. It is part of our guide to crosstabs explained: one-way and two-way tables, banners, and crosstab vs. pivot tables.
Key takeaways
- A one-way table shows how many respondents gave each answer to a single question, while a two-way table, or crosstab, shows how those answers differ between groups.
- A one-way table lists the count, percent and cumulative percent for each answer, and cumulative percent only makes sense when the answers have a natural order.
- Column percentages show what share of each group gave an answer, so put the group you want to compare across the top and read down each column.
- Software matters once a survey needs dozens of tables, because it builds them in one run and adds weighting and significance testing, such as a chi-square test.
What is a one-way table?
A one-way table shows the answers to a single categorical variable. A categorical variable is described by labels, such as hat color, dog breed, or an answer on a satisfaction scale. The table lists each answer once and says how many respondents chose it.
Three columns are common:
- Count (frequency): the number of respondents who gave that answer.
- Percent (relative frequency): the count divided by the total number of respondents.
- Cumulative percent: a running total of the percent column, from the first row down.
A tiny example: 10 men pick a hat color, and five choose red. The relative frequency for red is 5 / 10 = 0.50, or 50 percent. A one-way table is the tabular twin of a bar chart; both show the same numbers.
What does a one-way frequency table look like?
Here is a worked example. The numbers are illustrative, not real survey results. Imagine 200 customers rated overall satisfaction on a five-point scale.
| Satisfaction | Count | Percent | Cumulative percent |
|---|---|---|---|
| Very dissatisfied | 10 | 5% | 5% |
| Dissatisfied | 20 | 10% | 15% |
| Neutral | 50 | 25% | 40% |
| Satisfied | 80 | 40% | 80% |
| Very satisfied | 40 | 20% | 100% |
| Total | 200 | 100% |
The arithmetic: each percent is the count divided by 200 (10 / 200 = 5%, 20 / 200 = 10%, 50 / 200 = 25%, 80 / 200 = 40%, 40 / 200 = 20%). Each cumulative percent adds the next row to the one above (5 + 10 = 15, 15 + 25 = 40, 40 + 40 = 80, 80 + 20 = 100). The last value should reach 100; if it shows 99 or 101, rounding is the cause.
Read it like this: 60 percent of customers are satisfied or very satisfied (40 + 20), and 15 percent are dissatisfied or very dissatisfied. The cumulative column tells you that 40 percent rate their experience neutral or worse. Cumulative percent only makes sense when the answers have a natural order. For hat colors it means nothing.
How do you turn it into a two-way table?
Add a second variable. Suppose 80 of the 200 customers are new and 120 are existing. To keep the table small, group the five ratings into three: dissatisfied (the bottom two), neutral, and satisfied (the top two). The totals match the one-way table: 30 dissatisfied, 50 neutral, 120 satisfied.
| Satisfaction | New customers | Existing customers | Total |
|---|---|---|---|
| Dissatisfied | 20 | 10 | 30 |
| Neutral | 24 | 26 | 50 |
| Satisfied | 36 | 84 | 120 |
| Total | 80 | 120 | 200 |
This is a two-way table, also called a contingency table or cross-tabulation (see cross tabulation analysis for how to test and interpret one). The row and column totals, called margins, are the one-way tables for each variable.
Counts are hard to compare when groups differ in size, so convert them to percentages. You have two choices, and they answer different questions.
What do column percentages tell you?
Column percentages divide each cell by its column total, so each column adds to 100. They answer: among new customers, what share are satisfied?
| Satisfaction | New (n = 80) | Existing (n = 120) | Total (n = 200) |
|---|---|---|---|
| Dissatisfied | 25.0% | 8.3% | 15.0% |
| Neutral | 30.0% | 21.7% | 25.0% |
| Satisfied | 45.0% | 70.0% | 60.0% |
The arithmetic for the new column: 20 / 80 = 25%, 24 / 80 = 30%, 36 / 80 = 45%. For existing: 10 / 120 = 8.3%, 26 / 120 = 21.7%, 84 / 120 = 70%. Now the comparison is fair: 45 percent of new customers are satisfied against 70 percent of existing ones.
What do row percentages tell you?
Row percentages divide each cell by its row total, so each row adds to 100. They answer: of the people who are satisfied, what share are new customers?
| Satisfaction | New | Existing | Total |
|---|---|---|---|
| Dissatisfied (n = 30) | 66.7% | 33.3% | 100% |
| Neutral (n = 50) | 48.0% | 52.0% | 100% |
| Satisfied (n = 120) | 30.0% | 70.0% | 100% |
Check one row: 36 / 120 = 30% and 84 / 120 = 70%. These describe who makes up a segment, but they are easy to misread as a comparison. New customers are 40 percent of the sample (80 / 200), so their 30 percent share of the satisfied group says something different from the 45 percent of new customers who are satisfied.
A simple rule: put the group you want to compare across the top, and read the percentages down each column.
When should you use a one-way table and when a two-way table?
| One-way table | Two-way table | |
|---|---|---|
| Variables | One | Two |
| Question it answers | How many gave each answer? | Does the answer pattern differ between groups? |
| Typical use | Topline results, data checks, a first look | Segment comparisons, tracking by wave, subgroup analysis |
| Percent base | Total respondents | Column total, row total or grand total |
| Chart equivalent | Single bar chart | Clustered or stacked bar chart |
Start with a one-way table for every question. It shows the shape of the data and catches problems, such as an answer nobody chose. Move to a two-way table when a decision depends on a comparison.
To weigh a crosstab against a plain listing of responses, see crosstab vs. list: pros, cons and best uses.
Where does crosstab software fit?
Two tables are easy to make by hand, but a real survey is not two tables. With 40 questions and six audience groups you need dozens of tables, all built the same way. That is the job of crosstab software. Three things matter most.
- Volume. Software builds every question against the same columns (a banner) in one run.
- Weighting. If your sample has too many new customers compared with the customer base, you weight the data so each group counts in proportion. Weighted counts are rarely whole numbers, and the percentages must be calculated from the weighted figures.
- Significance testing. A gap between 45 and 70 percent might be real or might be sampling noise. For the illustrative table above, a chi-square test of independence gives a statistic of about 15.2 on 2 degrees of freedom, which corresponds to a p-value below 0.001. That is a strong signal that satisfaction differs by customer type, at least in this made-up sample.
Pivot tables hit limits here; the difference between crosstabs and pivot tables explains where.
Halo Reports, the mTab reporting product, builds crosstabs and banner tables with weighting and significance testing, and turns them into charts and interactive report pages. You can read more on the Halo Reports page. Insights teams can run wide, complex studies on one governed platform.
What are the common mistakes with one-way and two-way tables?
- Using cumulative percent on unordered answers. A running total of hat colors or brands has no meaning. Reserve it for ordered scales.
- Reading the wrong percentage. Row and column percentages are not interchangeable. Decide which question you are asking before you choose the base.
- Hiding the base. Always show n. A percentage without its base can mean 3 of 4 people or 300 of 400.
- Comparing tiny groups. In a group of 12, one respondent moves the result by more than 8 points. Combine categories or report counts instead.
- Reporting every gap. If you test dozens of cells at the 5 percent level, about one in 20 will look significant by chance. Focus on differences that are large, consistent and make sense.
What should you do next?
Take one survey question and build the one-way table first: counts, percent and, if the scale is ordered, cumulative percent. Then pick the one comparison your audience cares about, add that variable as columns, and read the column percentages. When you find yourself repeating this for every question, crosstab software saves the time, and you can then ask follow-ups in plain language.