Data Input
Columns represent one dimension (brands, segments, markets). Rows represent another (attributes, drivers, barriers). Cells must be numeric values.
Imagine Brand A scores 80 on every attribute while Brand B scores 40. In raw data, Brand A dominates the map simply because its numbers are bigger — not because it has a distinctive profile. The map shows size differences instead of shape differences. That is the big brand effect.
Standardisation removes this distortion so the map reveals what actually differentiates each column and row from the others — which is typically what you want.
- None (standard CA)
- Uses your data as-is. Best when all columns already have similar totals and no single column dominates. If your map looks like one cluster with an outlier far away, try switching to an equalised mode.
- Equalised (recommended)
- Removes the average level of each row and each column, keeping only the pattern of deviations. Think of it as asking: "Ignoring that Brand A scores higher overall, where does it score relatively higher or lower than its own average?" This is the safest general-purpose correction.
- Fully normalised
- Goes further — each row is rescaled so all rows have the same spread. Use this when some rows (attributes, drivers) have a much wider range of scores than others, causing them to dominate the map. This is the strongest correction; it forces every row to contribute equally.
Drag any label to reposition. Adjust axes, fonts, and markers above.
When to Use This & How to Read the Map
A correspondence map works whenever you have a grid of numbers where rows and columns represent two different things you want to compare. It compresses all the relationships in that grid into a single visual — showing you which rows are associated with which columns, which ones are similar to each other, and where the real differentiation lies. Any cross-tabulation with numeric cells is a candidate.