Hierarchical Card Sorting

What is Hierarchical Card Sorting?

Hierarchical Card Sorting (HCS) is a qualitative method to determine the categories people use when classifying problems/developments/interventions, what the content of those categories is and where they differ. By making this implicit knowledge more explicit, it is easier for this to be shared among the people involved. 

The method is a form of card sorting, or pile sorting, and can be used at several points during a research project. Before the start of the research it can be used to map the current situation, during the research for comparison with previous analyses, and after the research to map future activities. The outcome of HCS is a so-called treemap, a tree structure with branches.

How do you use HCS?
The method can be used for interaction with one person or a small group of people. The method breaks down into the following steps:

1. Identify the area of application or expertise you wish to explore. This could be a specific substantive topic (such as logistics in the big city), knowledge about projects in a particular area (such as AI initiatives/projects in the region), collaborations in a particular area (such as sustainability collectives or the stakeholders in a large programme), and so forth.
2. Make a list of occurrences (cases) in that application or area of expertise, for example: challenges that exist in that area, research questions prevalent in the area of expertise, projects carried out or planned in a particular area, collaborative initiatives in a particular area, and so on. Do not make the list too long (maximum 25 items) otherwise the follow-up steps will take a lot of time. Write each occurrence on a separate card.
3. Put all the cards in one pile. Ask participants an open-ended question about differences between the listed occurrences. This step is a warm-up exercise. The purpose is to make participants aware of the differences that undoubtedly exist between the occurrences written down on the cards. If this does not work straight away, two random cards can also be drawn from the deck and participants asked to compare them and name differences.
4. Ask participants to redistribute the stack into two piles, of any size, based on what they see as the main difference between all the occurrences written down on the cards. Emphasise that it is about what they consider to be the main difference.
5. To avoid the situation of a difference being ‘randomly’ named, the following question can be asked: “What difference does this difference make?” or “Why is this difference important?”.
6. Keep track of which cards have been placed in which pile (by numbering them, for example). Think of the first two stacks as the first branches emerging in a tree structure. You can also use this image by drawing a tree structure on a flipchart and writing the numbers at each branch. Write down what makes the difference between the two stacks.
7. Take one of the two stacks and repeat Steps 4 to 6. Do the same for the second stack. There are now four stacks. Repeat this until there is only one card left in each pile. Depending on what works best for the participants, choose either to work out each branch completely (depth) first, or to work it out layer by layer (width).
8. If it is not possible to arrive at just one remaining card, do not force the participants to continue but note down that no further difference was ascertained.
9. Document the results in a way that makes it clear how the different cases ended up in different categories and on the basis of what difference this was done. This can be in a tabular format or as a tree diagram.

The method depends on the participating individuals and their willingness to contribute. Mentioning differences between two stacks of cards can sometimes be difficult, for example because it was stated that there is no difference, or a ‘superficial’ difference was chosen or that several differences were mentioned. It is important to ask what the meaning of the differences made by the participants is.

Once the tree structure has been mapped, additional questions can be asked at each branching of the tree structure, for example “which of these two groups lends itself best to intervention in the coming months?” or “what can we contribute most to in the coming period?”. The choices made at each branching can then be used to make a ranking of which category correlates ‘best’ to the question asked.

What is the origin of HCS?
The HCS method was originally designed by Rick Davies in 1993, as was the MSC method. During a large and complex development project in Bangladesh, he ran into the problem of monitoring and evaluating the project. The large scale of the project – tens of thousands of participants in hundreds of villages – and the open-ended design of the project made it difficult to use a traditional method. Getting everyone on board with a set of indicators was impossible, also because of the great diversity of activities and many conflicting interests.