Start with the rows, columns and question
A cross-tabulation, often called a cross-tab, organizes observations by two categorical variables. One variable labels the rows and the other labels the columns. Each cell counts observations that fit both labels. For example, a survey might record whether a workshop participant attended online or in person, and whether that participant prefers updates by email or text.
Before interpreting a cross-tab, identify what one observation represents. Is each entry a person, a visit or a response? Then read the row and column labels, including any category called “Other,” “Unknown” or “Missing.” Finally, put the table’s purpose into words. In this example, the question is: Does communication preference differ between online and in-person attendees?
Read a cell, then check the totals
Consider this invented table for practice. Of 60 online attendees, 36 prefer email and 24 prefer text. Of 60 in-person attendees, 18 prefer email and 42 prefer text. The email column totals 54 people; the text column totals 66. The grand total is 120. A cell such as “36” means that 36 people are both online attendees and email-preferring respondents. It does not mean that 36% of attendees prefer email.
Totals provide a quick check on your reading. The two online cells add to the online row total of 60, and the two in-person cells add to 60. The column totals add to 120. If a table includes missing responses, its displayed cells may not add to the number of people originally surveyed. Look for a note explaining whether incomplete records were excluded or placed in a separate category.
Choose the denominator that answers your question
Counts show how many observations fall in each cell, but percentages are usually more useful for comparing groups of different sizes. To ask, “Among online attendees, what share prefers email?” divide the online-and-email count, 36, by the online row total, 60. That is 60%. Among in-person attendees, 18 out of 60, or 30%, prefer email. These are row percentages: each compares people within the same attendance group.
A column percentage answers a different question. Among everyone who prefers email, 36 of 54 attended online, or about 67%. That figure describes the composition of the email-preferring group; it does not say that 67% of online attendees prefer email. Percentages of the entire table answer yet another question: 36 of all 120 respondents, or 30%, are both online attendees and email-preferring respondents. Always name the group after “out of” before interpreting a percentage.
Some cross-tabs print counts and percentages together. Check the heading or footnote to see whether those percentages are calculated by row, by column or from the grand total. If no denominator is identified, use the counts and totals to calculate the percentage you need rather than guessing.
Describe the pattern without stretching the conclusion
In the practice table, email preference is more common among online attendees than among in-person attendees: 60% versus 30%. You could describe the difference as 30 percentage points. Text preference shows the reverse pattern, at 40% among online attendees and 70% among in-person attendees. This is an association in the hypothetical responses, not proof that attendance mode causes anyone to prefer a particular channel.
The table alone cannot explain why the groups differ. Participants might vary in age, access to devices or reasons for attending, for example. Nor does the pattern automatically apply beyond the people represented. To judge how broadly it might apply, you would need to know how participants were selected, how the questions were worded and how missing answers were handled. With very small groups, a difference of only a few people can also produce a large-looking percentage gap.
Use a short reading routine
When facing a new cross-tab, first identify the unit being counted and both variables. Next, read one cell aloud as a combination of its row and column labels. Check the row, column and grand totals. Then choose a denominator based on the question: use row percentages to compare outcomes within row groups, or column percentages to compare the makeup of column groups. If group sizes differ, compare percentages alongside the underlying counts.
Finish with a restrained sentence that states the observed pattern, the groups compared and any important limitation. For the invented workshop table, that might be: “Email preference was higher among online attendees than in-person attendees in these responses, though the table does not explain the difference.” This keeps the cross-tab useful without asking it to answer a question it cannot resolve.
