Banner tables
Significance Letters in Crosstabs, Explained
The letters on a market-research crosstab are column-proportions tests: every banner column gets a letter, and a cell shows the letters of the columns it is significantly higher than — uppercase at the stricter confidence level (usually 95%), lowercase at the looser one (usually 90%).
Reviewed by the crosstabs.com methods team · Last updated
Run this on your own data — free, no signup
Upload a CSV or XLSX. Everything runs in your browser; your file never leaves your device.
Open the workspace →Reading a lettered table
Suppose your banner has Gender (Male = A, Female = B) and Age (18–34 = C, 35–54 = D, 55+ = E). In the “Aware of brand” row, the Male column shows 42% B. That single letter says: among men, awareness (42%) is statistically significantly higher than among women, at the 95% confidence level.
If it showed 42% b (lowercase), the difference clears only the 90% bar — suggestive, not conclusive. No letter means no significant difference against any other column in that row.
How the test works
For each row, every pair of banner columns is compared with a two-proportion z-test on the column percentages. Because a table with five banner columns runs ten comparisons per row, tools apply a multiple-comparison correction — SPSS Custom Tables and crosstabs.com use Bonferroni (the significance threshold is divided by the number of pairs); Q and Displayr default to the False Discovery Rate.
Columns with small bases (under ~30) are excluded, and weighted data is tested on weighted counts with the usual caveat that the weights are treated as frequency weights.
Run lettered tables free
This notation has historically lived behind expensive software — SPSS Custom Tables, Q (~$3,259/yr), Displayr, WinCross — or paid survey-platform analytics add-ons. crosstabs.com computes the same letters free: open a CSV or XLSX in the workspace, build your crosstab, and switch on Letters in the display toggles. The letters survive CSV and Excel export, so your client deliverable keeps the notation.
How to interpret it
Rule of thumb
A letter answers one precise question: “is this column's percentage higher than that column's, beyond what sampling noise explains?” It says nothing about effect size — pair letters with the percentages themselves, and Cramér's V for the table overall.
Frequently asked questions
- What do the letters on a crosstab mean?
- Each banner column is assigned a letter (A, B, C…). Inside the table, a cell's letters list the columns whose percentage it is significantly higher than. For example, '34% B' under column A means column A's 34% is significantly higher than the same row's percentage in column B.
- Why are some letters uppercase and some lowercase?
- Most tools run the tests at two confidence levels. Uppercase letters mark differences significant at the stricter level (typically 95%), lowercase at the looser level (typically 90%). A lowercase letter is a weaker signal worth noting, not a confirmed difference.
- What statistical test is behind the letters?
- A pairwise two-proportion z-test between every pair of banner columns within each row, usually with a Bonferroni correction because many comparisons are made at once. SPSS Custom Tables, Q, Displayr, WinCross, and crosstabs.com all follow this approach.
- Why do some columns show no letters at all?
- Columns with a small base (commonly under 30 respondents) are excluded from testing because the z-test is unreliable at small sample sizes. crosstabs.com marks these columns with an asterisk.
References & further reading
- IBM SPSS Statistics — Custom Tables: comparing column proportions (z-tests with Bonferroni adjustment)
- Agresti, A. (2013). Categorical Data Analysis, 3rd ed.
Try it on your own data — free, no signup
Upload a CSV or XLSX. Everything runs in your browser; your file never leaves your device.
Open the workspace →