Fieldwork has closed, the data file has landed and a stakeholder wants to know whether younger customers feel differently from older ones. Survey crosstabs answer that, question by question, for every group the study cares about. Our guide to what a crosstab is and how one-way and two-way tables work covers the table itself. This article covers the practitioner’s workflow: how research and insights teams crosstab a survey end to end, from banner to final check.
Key takeaways
- Survey crosstabs show every question broken down by the subgroups that matter, so you can compare how groups answered.
- A topline gives total-sample results; a crosstab gives the same results by subgroup, and its total column should match the topline.
- Bases, weights, nets and question type decide what each percentage means, so set them first.
- Significance letters tell you which differences between columns are larger than sampling noise.
- QA every run against the questionnaire routing and the topline.
What are survey crosstabs and why do teams rely on them?
A survey crosstab takes one question, such as purchase intent, and shows the answers for the total sample and for each banner group: age bands, regions, customer segments, brand users. A full set repeats that for every question against the same banner.
They are the workhorse of survey analysis because they answer the question stakeholders actually ask: who thinks what? Most findings in a survey report start as a crosstab cell that differs from its neighbor. Our wider guide on how to analyze survey data shows where crosstabs sit among the other methods.
What is the difference between toplines and crosstabs?
Toplines and crosstabs come from the same data. They differ in how far they break it down.
| Topline | Crosstab | |
|---|---|---|
| What it shows | Results for the total sample, question by question | The same results split by subgroup |
| Columns | One (total) | Total plus every banner point |
| Typical use | First read of the data, quick headline numbers | Comparing groups, finding the story, significance testing |
| Check between them | The crosstab’s total column must equal the topline |
How do you crosstab survey data, step by step?
Most of how to crosstab survey data well is doing the setup in the right order.
- Define the questions the study must answer. The decisions the research supports tell you which groups to compare.
- Choose banner points. The banner is the set of columns: typically demographics, region, customer type and key behaviors. Our guide to designing banner tables for market research covers layout and grouping.
- Clean the data and check the respondent count. Remove speeders, duplicates and failed attention checks, then confirm the completes.
- Set bases and filters per question. The base is the people who were asked the question. If only current users saw Q12, the Q12 base is current users.
- Apply weights. Weight to known population or customer targets and show both unweighted and weighted bases.
- Handle each question type.
- Single-response: column percentages that sum to 100%.
- Multiple-response: each option is a percentage of respondents, so columns can sum to more than 100%. Our article on filtering and analyzing multiple-response questions goes deeper.
- Grids and rating scales: a table per grid row or a summary table, with top-2-box, bottom-2-box and a mean.
- Numeric questions: means and medians, plus banded ranges.
- Open ends: code the verbatims into themes first, then crosstab the codes like a multiple-response question.
- Add nets and summary rows. Nets combine answers (any of three brands, total satisfied) and count each respondent once.
- Add significance testing. Choose the confidence level and which columns to compare, usually within each banner group.
- QA the tables. Use the checklist below.
- Deliver and chart. Send the tables with base notes, then chart the differences that drive the story.
What does a survey crosstab look like? A worked example
The numbers below are illustrative. A company surveyed 800 customers and asked how likely they are to recommend it on a 0 to 10 scale. Answers are collapsed into promoters (9 to 10), passives (7 to 8) and detractors (0 to 6). The data are weighted so that region splits 50/50 and customer type 40/60 new to existing. Letters mark significant differences at 95% confidence within each banner group.
| Total | North (A) | South (B) | New (C) | Existing (D) | |
|---|---|---|---|---|---|
| Unweighted base | 800 | 440 | 360 | 280 | 520 |
| Weighted base | 800 | 400 | 400 | 320 | 480 |
| Promoters (9–10) | 38% | 42% B | 34% | 35% | 40% |
| Passives (7–8) | 36% | 34% | 38% | 36% | 36% |
| Detractors (0–6) | 26% | 24% | 28% | 29% | 24% |
| Net Promoter Score | 12 | 18 | 6 | 6 | 16 |
How the numbers reconcile
North was over-represented (440 completes) and South under-represented (360), so weighting brings each to 400. The total column reconciles with each group: North’s 42% of 400 is 168 promoters and South’s 34% of 400 is 136, and 168 + 136 = 304, which is 38% of 800. New and existing give the same total: 35% of 320 is 112, 40% of 480 is 192, and 112 + 192 = 304.
Net Promoter Score is promoters minus detractors: 38 − 26 = 12 for the total, 42 − 24 = 18 for North, 35 − 29 = 6 for new customers.
Reading the significance letter
The “B” next to North’s 42% means it is significantly higher than South’s 34%. A two-proportion test on the unweighted bases gives z ≈ 2.31, above the 1.96 cutoff for 95% confidence. The five-point gap in promoters between new (35%) and existing customers (40%) gives z ≈ 1.39, so it carries no letter. For more on interpreting letters and bases, see our guide on how to read a crosstab.
How do you QA survey crosstabs?
We see the same few errors in almost every first run. This list catches most of them.
- Bases match the questionnaire routing: every filtered question shows only the people who were asked it.
- Total columns reconcile with the toplines, question by question.
- Weights sum to the sample size, and weighted bases match the targets.
- Nets do not double count: a respondent who picked two brands in a net counts once.
- Mutually exclusive banner columns add up to the total, and single-response columns sum to 100%.
- Small bases are flagged, and significance letters compare the intended columns.
How do you crosstab a survey across tracker waves?
In a tracker, you crosstab a survey the same way every wave, then compare waves. Keep banner definitions, nets, weighting targets and question wording fixed, or the change you report may be a change in method. Add wave as a banner, or put waves side by side within each subgroup, and test wave-on-wave differences. Our article on comparing past and present survey results lists what else to check before calling a trend.
Should you use a spreadsheet or a dedicated crosstab tool for survey data?
A spreadsheet works for a short survey with a few banner points; our guide to building a crosstab in Excel shows how. It gets hard once you need multiple-response questions, weights, nets and testing across hundreds of tables, each set up by hand.
A dedicated crosstab tool for survey data treats bases, weights and testing as settings and reruns the whole set when the data changes. Our overview of what crosstab software does and how to choose it covers the features to compare. If you already have the data, mTab provides crosstabs and banner tables with weighting and significance testing, along with charts and interactive report pages.
What are the common mistakes with survey crosstabs?
- Using the total sample as the base for a filtered question. The percentages look low and mean nothing.
- Testing on weighted bases as if they were real respondents. Weighting changes a group’s size on paper, not the evidence behind it. Test on the unweighted (or effective) base.
- Building nets by adding percentages. Two brands at 30% each do not make a 60% net if some people chose both.
- Adding banner points nobody asked for. Every extra column multiplies the tables and the chance differences.
- Changing definitions between waves. A new age break or weighting target creates a trend that is not in the market.
What should you do next?
Before your next study closes, write the banner and each question’s base into a one-page table spec and agree it with the people who will use the results. Then crosstab survey data against that spec, QA it and chart only the differences that carry a letter. For teams running wide studies and trackers across markets, mTab offers one governed platform for insights and research teams.