Listen to the audio version on Spotify and Apple Podcasts.
Useless to useful: how to cluster
In my latest blog (LinkedIn newsletter), I explained the importance of clusters in research processes and why getting projects into clusters (classifications, groupings, patterns, themes) is so important.
In this edition I will zoom in on types of clusters. Next time, we will look at how you know if your approach is right.
To start with, an important issue when getting your clusters right is this: There is no right or wrong. There is only more useful and less useful. This sometimes makes decisions difficult, especially with team members who are (only) familiar with certain types of clusters and do not fully understand the reasons beyond getting some kind of order into your projects.
What am I talking about here exactly? Imagine trying to understand and analyze a group of 100 innovative projects that are run by 100 different organizations and come from 100 different countries. There is a clustering approach that always works and that everyone is familiar with: grouping the project names or organization names by the alphabet. So A–F, G–M, N–S, and T–Z, as you will know it from registration desks at conferences or from grouping students for seminars.
Useless clustering
Everyone is familiar with alphabetic clustering, and it is as systematic as it can be, so is this a good way of clustering? Most of the time, it is useless, since projects in one cluster have nothing in common and share with their “peers” in the cluster just as much as with those in the other clusters.
In other words, the groupings are close to random and you cannot read anything out of any cluster, or into them. Actually, it is even worse than that, since you will have some uncontrollable bias in your clusters. Names of projects and organizations from Arab countries, for example, very often start with A, and some organizations that choose their own names often try to be first in every list. So, coming back to the useful vs. right argument: There is nothing wrong with it, but I have never come upon any real use case for such clusters.
Navigating organizational names
And, in our context of understanding innovation processes, it is not even as systematic as it might seem. Take names of organizations. They may come as full names, or as abbreviations, or even as acronyms. They may appear in their native language and in their English translation. And when they are part of a larger organization, it may be the larger organization first (like a university) or the unit/department first (like an institute).
Clustering by region
So, let us move on to something that is usually more useful: regions. These clusters are useful not only because of the systematic and well-known method. Projects in the same regional clusters are often those that decision-makers consider first, or exclusively, because their peer group consists only of projects from neighbouring countries or from “their” region.
But even clustering by region is not as straightforward as it might seem from the outside:
- Projects are often domiciled in one country but implemented in another. Often, they are implemented in many countries from different regions. So, you have to decide: region of domicile or region(s) of implementation?
- Clustering by continents looks like a clear approach, but what about Turkey, Russia, Georgia, and Armenia? There is no clear line that divides Europe and Asia.
- For many purposes, it makes sense to subdivide regions, like clustering by Central & Eastern Europe or Northern Europe. But what about the rest of Europe? Usually, Southern Europe (putting everything from Turkey to Portugal in one cluster) does not make sense.
- There are also trans-continental clusters like MENA (Middle-East and Northern Africa) that can make sense for your clustering, or clusters that bring projects from countries with similar income and development levels (like OECD) into one cluster. Sometimes membership in political or economic unions (like the EU or ASEAN) are good clusters.
- Sometimes, especially when it comes to technology, the regional approach becomes blurred or useless. Does it make sense to put all projects from ICT multinationals into the cluster “North America” because most of them have their home base there, whereas the app is available for download in every country? Sometimes the clustering option “global” makes sense, sometimes not – depending on your use case.
Types of disability
Since the Zero Project works with projects related to disability, inclusion, and accessibility, getting projects clustered by types of disability is deeply ingrained in our research methodology. And again, it is not as easy as it may look from the outside.
There is no commonly accepted and at the same time all-encompassing definition or typology of disabilities. There is the Washington Group and its definitions, and the Zero Project is using them. But the Washington Group definitions do not provide any guidelines when it comes to persons with intellectual or psychosocial disabilities.
Washington Group typologies
What we at the Zero Project did is this: We categorize every project by the Washington Group typologies, but in parallel we have developed our own set of disability types, based on their frequency within nominations. This is a list that contains types of disabilities like autism, albinism, dyslexia, Amyotrophic Lateral Sclerosis (ALS), or Post-Traumatic Stress Disorder (PTSD). They sometimes overlap, and they keep evolving. Most importantly, there are projects that work for all persons with visual impairments, but some only work for those who have impairments and are not fully blind (like technologies enhancing remaining eyesight). When it comes to physical disabilities, some projects work for all, but some only for those using a wheelchair.
Mainly or exclusively
There is another hidden challenge here. When you research and cluster, you might:
- Mission 1 – be interested in projects that focus on a certain type of disability, like those that support Deaf persons.
- Or, Mission 2 – look at cross-cutting projects that also support Deaf persons, like advocacy campaigns or accessibility regulations.
If you are on Mission 1, you do not want to get a broad advocacy campaign into your cluster “Deaf persons”. If you are on Mission 2 and you only get “Deaf person-only” projects into the cluster you want to build, you will miss out on many projects that also support Deaf persons.
At the Zero Project we have defined both, so we can cluster for Mission 1 or Mission 2. But this is complex to understand and to work with, even for our experienced team, so we are still not really happy with the practical work in this regard.
Themes and topics
There is every reason to cluster projects according to their themes. For the Zero Project, this is – next to region – the most obvious approach, since our whole research process is structured according to four main topics, which are based on the CRPD: employment, accessibility, independent living and political participation, and education. In a four-year-cycle, our research focuses on one of these.
In 2019, we added Assistive Technology as a theme that we research every year. For every topic we address, we identify sub-topics, which serve the purpose of structuring our selection process, but also of providing guidance to our network what kind of innovative solutions may be nominated. “Independent living” is clearly too abstract for innovators to understand whether their nomination will be considered by the Zero Project, so we break it down into community-based approaches, deinstitutionalization, and supported decision-making, among others.
Sectors of society
It is not only crucial to get the clustering right for a project, but also for the organization behind it. At the Zero Project we work with three main clusters to categorize the sector from which the project comes:
- Civil society (OPDs, NGOs, many non-profit service providers, foundations)
- Business sector (including umbrella organizations)
- Public sector (including also government-related agencies like regulatory bodies or the judicial system)
These clusters work well for basic purposes, but sometimes they need additional ones, or “sub-sectors”, depending on the research angle. Hospitals, universities and schools, museums, and the media usually behave differently from other organizations within their sector, so they might need a separate cluster. In museums, for example, reaching and appealing to customers is key, even when they are run and funded by a public sector authority that usually work with regulations, so do not have a lot in common with other projects in the cluster “public sector”.
Also, supranational organizations like UN agencies, as well as international development funding banks or international standardization organizations, will often need their own cluster. If you cluster within one country, municipalities (city or other local authorities) and districts (provinces, federal states) become important.
Other types for other purposes
There is a large number of other types of clusters that can make sense, depending on your research angle:
- If you are an investor or grant giver, you want to understand the status of development, the scaling strategy, and the long-term income model.
- If you are a human rights expert, you may want to understand the relation to gender equality, indigenous peoples or other minorities, or whether a relation to crises of all sorts exists.
- If you are mainly interested in technology, you will be mainly interested in how Artificial Intelligence is used, 3D printing, or whether it is useful for orientation, training, or enabling communication.
At the Zero Project we have been evolving our cluster strategies ever since we started researching good practices and policies; and we continue to do so to reflect our changing world.
In the next edition of this newsletter/blog we will investigate how to approach the process of clustering in practice. In the meantime, please feel free to share your view and experience with clusters in your own work!
Read more