Market research method
What is a banner table? Reading and building survey banners
A banner table is a survey crosstab that shows one question down the side (the stub) against several breakdown variables across the top (the banner) — for example Total, gender, age and region side by side. Each column shows the percentage of that subgroup giving each answer, its base, and letters marking columns that are significantly higher.
How results are checked
Named statistical review pending. Selected automated reference cases are published on the validation page.Build a banner table — free, no signup
Open your CSV, Excel, SPSS or Stata file in the workspace, choose All tools → Bulk tables, pick the banner variables and the row questions, and export the lettered tables to Excel. Everything runs in your browser.
Build a banner table →The parts of a banner table
Stub. The rows: the answers to one question (or a net of answers, such as “Top 2 box”). In a tab book every question gets its own page against the same banner.
Banner and banner points. The columns: one or more breakdown variables placed side by side. Each column is a banner point — Female, Male, 18–34, North and so on. The points within one banner variable should not overlap; different banner variables are independent cuts of the same respondents.
Total column and base row. Total comes first and covers everyone who answered the question. The base row prints each column's n, so a reader sees at once when a percentage rests on too few people.
Column percentages and significance letters. Each cell is a column percentage, so the columns compare like with like. Every column carries a letter; a letter printed in a cell names a column in the same banner that this cell is significantly higher than.
The method
Formula
z = (p₁ − p₂) / √( p̄(1 − p̄)(1/n₁ + 1/n₂) ), p̄ = (x₁ + x₂) / (n₁ + n₂)
where
- p₁, p₂
- the column percentages (as proportions) of the two banner points compared
- n₁, n₂
- the two column bases
- x₁, x₂
- the counts giving that answer in each column
- letters
- Crosstabs adjusts each p-value with Bonferroni across the pairs in one banner, prints an upper-case letter at 95% and lower case at 90%, and skips columns with a base under 30
Worked example
The synthetic snack-brand tracker that the free research tools load as sample data has a 0–10 likelihood-to-recommend question. Recode it into NPS groups and run it against three banners — gender, age group and region — exactly as the workspace's Bulk tables view does:
| NPS group | Total | Female (A) | Male (B) | Non-binary (C) | 18-34 (D) | 35-54 (E) | 55+ (F) | East (G) | North (H) | South (I) | West (J) |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Base (n) | 600 | 306 | 276 | 18 | 203 | 216 | 181 | 132 | 133 | 166 | 169 |
| Detractor (0–6) | 39.5% | 37.6% | 42.0% | 33.3% | 42.9% | 35.6% | 40.3% | 43.9% | 46.6% i | 31.9% | 37.9% |
| Passive (7–8) | 35.5% | 33.0% | 37.0% | 55.6% | 34.0% | 37.5% | 34.8% | 35.6% | 29.3% | 33.7% | 42.0% |
| Promoter (9–10) | 25.0% | 29.4% B | 21.0% | 11.1% | 23.2% | 26.9% | 24.9% | 20.5% | 24.1% | 34.3% GJ | 20.1% |
Synthetic snack-brand tracker · 600 respondents · the same data as “Use sample data”. Upper-case letters: higher than that column at 95%; lower case: at 90%; Bonferroni-adjusted within each banner.
Reading it. 29.4% of women are promoters against 21.0% of men, and the B in the Female cell says the gap is significant at 95%. In the region banner, South has 34.3% promoters and carries GJ: it beats East and West at 95% after the Bonferroni adjustment. North's 46.6% detractors carry a lower-case i, a weaker difference that only clears 90%.
Non-binary (n = 18) is shown but not tested: below a base of 30 a column takes part in no comparison, because a percentage that rests on a handful of people moves a lot from sample to sample. The age banner has no letters at all — its chi-square p is 0.66, so the three age groups recommend the brand at similar rates.
Using it well
When to use it
Use it when
- Use: Reporting a survey to stakeholders who want every question broken down by the same audience cuts.
- Use: Scanning many questions quickly for where subgroups differ — the letters point to the cells worth a closer look.
- Use: Producing a tab book: one banner, every question in the questionnaire, exported to Excel.
Not the right tool when
- Avoid: Drawing conclusions from columns with small bases; show them, but do not test or headline them.
- Avoid: Treating every letter as a finding: dozens of questions × many banner points produce some false positives even with an adjustment.
- Avoid: Explaining why a difference exists. A banner shows where groups differ; a driver model or a follow-up study explains it.
How to interpret it
Rule of thumb
Read across a banner, never between banners: a letter compares columns within one banner variable only. Check the base row before the percentages, and treat a lower-case letter as a lead to check, not a result.
Planning a banner
Start from the decisions the study supports. A brand tracker usually needs demographics plus the behavioural cuts the client acts on — users versus non-users, heavy versus light buyers, region. Keep the banner to what fits on one landscape page; a second banner is cheaper than an unreadable one.
Combine small groups before fielding, not after: if a banner point will fall under about 30 completes, net it with a neighbour (for example 55–64 and 65+ into 55+) or flag it as indicative. Keep the same banner across waves so trend tables line up.
Building one in Crosstabs
Open the data file in the workspace (CSV, Excel, SPSS .sav, Stata, SAS, R, Parquet or JSON — it stays in your browser). Choose All tools → Bulk tables, pick the banner variables (columns) and the row questions, and run. Each row question becomes one table with a Total column and the banners side by side; letters continue across banners (A–C, D–F, …) but each banner's columns are only compared with each other.
Excel export writes one workbook with a contents sheet. For a versioned tab book with saved multibanners, filters, weights and export recipes, use the Tab book view.
Questions
Frequently asked questions
What is the difference between a banner and a stub?
The banner is the set of columns — the breakdown variables such as gender, age and region placed side by side. The stub is the set of rows — the answers to the question being analysed. In a tab book the banner stays the same and the stub changes from page to page.
What are banner points?
Banner points are the individual columns of a banner: Female, Male, 18–34, North and so on. Each has its own base, and each gets a letter that significance tests refer to.
What do the letters in a banner table mean?
A letter in a cell names another column in the same banner that this cell's percentage is significantly higher than. In Crosstabs an upper-case letter means significant at 95% and a lower-case letter at 90%, after a Bonferroni adjustment across the pairs in that banner.
Why are some columns not tested?
Columns with a base under 30 are shown but left out of the significance tests: no other column is compared with them and they print no letters. Combine the group with a neighbour or report it as indicative.
Is a banner table the same as a crosstab?
It is a crosstab with several column variables placed side by side and a Total column first. A plain crosstab has one row variable and one column variable; a banner table repeats the row question against every banner variable in one view.
Next steps
Build a banner table — free, no signup
Open your CSV, Excel, SPSS or Stata file in the workspace, choose All tools → Bulk tables, pick the banner variables and the row questions, and export the lettered tables to Excel. Everything runs in your browser.
Build a banner table →