A matter of scale
The anthropologist Robin Dunbar posited that the number of social relationships which any individual can sustainable create is between 100 and 250, with a typical average of 150. Studies of social groups repeatedly find evidence of this, leading to Dunbar’s Number being recognised as a constant.
In the same way, data and analytics functions need to have reached a certain scale before the concept of squads and tribes becomes relevant and viable. This is a consequence of the internal logic of the approach itself and also of human nature. A small unit of, say, 10 to 30 practitioners will be able to operate in a stable, self-organising way without any formal overlay. Similarly, the mathematics of squads and tribes demands a minimum of around 20 practitioners in order to be meaningful.
Defining the structure
While the concept of squads and tribes has a fixed meaning, the definitions of what they do will vary by organisation. This is best set out through a process of discovery and mapping, such as that carried out by a national grocery chain. After management learned about this working practice and decided to adopt it, it spent time in workshops identify all of the activities which the data and analytics function was carrying out.
These were then grouped logically as product areas under a product manager. Each of these products is a tribe, since it involves practitioners from across the function who routinely carry out these tasks, for example, building data models, segmenting customers, defining KPIs, etc. Tribes are usually limited to under 100 members.
To respond to specific projects required by stakeholders, multi-disciplinary teams were needed – these form the squads. Each of these faces off to a specific stakeholder in the business and is typically made up of between three and eight practitioners. Squads can be self-selecting under the supervision of a scrum master to ensure balance and fairness. At the completion of each project, the squad breaks up and returns to its tribe.
Communities of practice are supra-tribal groups who share an interest, skill set or desire to learn, for example, specific coding languages or analytics technologies. Some organisations have both chapters – groups of specialists with common responsibilities – and guilds – informal groups with common interests – although this is generally the preserve of very large-scale data and analytics functions.
Advantages of squads and tribes
Adopting this approach is generally seen as a positive because of the way it mirrors the agile sprint cycle, creates cross-functional working and knowledge-sharing, while avoiding fixed size teams which lack the ability to flex their resources according to demand.
Practitioners can welcome this model because it is based on trust, rather than command and control. In large functions, it creates connections at multiple levels while also maintaining a sense of identity. Exposure to new and different projects keeps practitioners interested and engaged.
Challenges to squads and tribes
This approach is not the same as an organisational chart and is not about the specific roles involved. As a consequence, it can seem too loose for hierarchical organisations and can be confusing given the multiple memberships which any one individual can have (squad, tribe, community of practice or guild).
Since the core task of analytics managers becomes more soft skills and development focused, it can prove a difficult transition to make from within a conventional structure. Stakeholders may also be concerned by the loss of a permanent relationship with specific practitioners, since squad members are likely to vary over time. Autonomy for practitioners can also seem excessive if it extends to how all resources and time are allocated. Diversity may also not be present in a system of self-selection.
Keeping agile
To get the best out of squads and tribes as a model, assuming there is sufficient scale to make it feasible, it is also important to subject the approach to its own continuous improvement processes. The model is intended to be flexible and adaptable, rather than a fixed solution to how data and analytics teams should work.
