My Affinity Chart Method

My Affinity Method

When performing research via surveys, it can be tricky to categorize free text responses.

One of the most useful techniques I’ve developed follows the general proceedure of creating an Affinity Diagram.

Three Rules

  • Do NOT make category assumptions. It doesn’t matter if you know what the themes will be.
  • Record an ID to get back to a response. It saves a lot of time and headaches to record a unique ID (or email address or something) so you can find a person later for following up.
  • Do NOT group your groups until you’re done processing. It doesn’t take that long to do, which is all the more reason to just wait until you’ve finished going through the raw data.

The Process

This process assumes you’ve loaded up a bunch of raw survey data into some flavor of spreadsheet app.

Step 1: Set Up Column Focus

  • Decide which freeform response you are going to categorize, and hide every column except for the question and an identifier column (UUID, email address, or generate your own).
  • Sort the responses in reverse alphabetical (Z-A, yeah) order. Now you don’t have to scroll past a bunch of empty cells to get to words.

Step 2: Start a New Blank Spreadsheet Workbook Thing This new sheet will be where the categorization happens. I tend to throw everything into a single sheet, and then add additional sheets as needed (the tabs on the bottom when looking at Google Drive).

The first row will contain headers for each of the general themes that will emrge. Each header should have two columns between them, because the data rows will have some text and the identifier.

Step 3: Initial Reading & Classification Move down, row-by-row, and read the responses. Find the theme(s) of the text, and record some or all of the text under the appropriate column, along with the unique id so that you can find the response later.

Here’s an example:

The pizza was great! I love all the unique topping choices, and the music was rad. The ponly thing is that the kid working the register was a little rude.

| Toppings        |    |   | Music               |    |   | Staff |    |
| Loved selection | id |   | "the music was rad" | id |   | Rude  | id |
|                 |    |   |                     |    |   |       |    |

Step 4: Summarization After breaking responses into groups, it becomes easier to quantify the qualitative data. You could split each theme into positive & negative, or take it further and group respondents in one theme by answers to another question to find other insights.

Check out my other articles, or sign up for my newsletter:

No spam. Just code stuff, unsolicited music recommendations, etc.