Surface Map

Syntagm > Design for Usability > Resources > Card Sorting | Surface Map

Surface maps can provide a simple visual analysis of card sort results. Start with a spreadsheet - either created by hand or output automatically by SynCaps. A partial sample is illustrated below (this can be obtained by processing the data file that is included with the SynCaps executable). The values that appear in each cell are the counts of the number of times each pair of items appeared in the same group.

Proximity Matrix from SynCaps (first 5 columns only)

In Microsoft Excel, select the whole matrix and use the Insert > Chart menu to create a surface chart of the contour sub-type. See the screenshot below:

Insert Chart Dialog from Microsoft Excel 2003

Now, just click Finish. To get the result as it appears below, there are only two more steps:

1. Double-click anywhere in the right axis. The Format Axis dialog should appear. Select the Scale tab and check "Series in reverse order". This will put the Y axis in the same order as the X.
2. Double-click anywhere in the legend to the right. This will display the Format Legend dialog. Click on the Scale tab and change the Maximum and Major unit fields appropriately (to 12 and 3, respectively, in this case). When you dismiss the dialog you can then also drag the legend elsewhere on the chart if you like.

The surface map will now look like this:

Surface Map (rows and columns in arbitrary order)

It helps us to spot closely related items fairly easily, but is not useful for grouping yet. The next step is to move related items closer together in the spreadsheet. (You can skip this step for SynCaps version 1.1 and later since the matrix is output in dendogram order.) Do this by cutting and pasting (or dragging) rows and columns as required. You will need to create a new contour map when you have finished rearranging the rows and columns - Excel loses track in the existing map.

In our example, Beaujolais, Merlot, Claret, Syrah and Cabernet Sauvignon all have pretty high counts between each pair - mostly tens. So in the spreadsheet, we group those rows and columns together (just decide whether you prefer to move the rows or the columns and then adjust the other to match). Similarly, Chardonnay, Riesling, Muscat and White Zinfandel share some pretty high pair counts, with Pinot Grigio trailing slightly behind. Finally, Champagne and Cava have a high pair-count too, so we group them together at the end. The only other adjustment is to make the diagonal of the matrix (originally all zeroes) blend in with its neighbours. So, for example, the Beaujolais-Beaujolais cell becomes ten since that is the value of many of the surrounding pairs.The revised spreadsheet and contour map now look like this:

Reordered Proximity Matrix from SynCaps (first 5 columns only)

Reordered Surface Map

By reordering the columns and rows in the matrix the contour map now shows fairly clearly three distinct groups. The "strongest", in the top left, is the red wines, the whites are shown in the middle, and the small two-element sparkling wines group is in the bottom right. The "antlers" that appear to be radiating from the central group are due to Pinot Grigio - a white wine that was not very familiar to participants, as can be seen by the pair scores it has with many of the red wines.