Essential Scrum – Scrum Team Structures (6)
Feature Teams versus Component Teams
- Feature Team – A feature team is a cross-functional and cross-component team that can pull end-customer features from the product backlog and complete them.
- Component Team – A component team, on the other hand, focuses on the development of a component or subsystem that can be used to create only part of an end-customer feature. Component teams are sometimes referred to as asset or subsystem teams. Often a community of practice made up of people with a similar specialty skill also functions like a component team. On these teams, all members likely report to the same functional manager and might operate as a shared, centralized resource to other teams. One example might be the centralized UX department that creates UI designs for other teams.
Scrum favors feature teams.
Components Teams Working on Multiple Products
In this example, there is no feature team that works on a complete product backlog item. Instead, a feature is selected from the top of the product backlog and is split into its component-level pieces. This splitting is done either collectively by the members of the component Scrum teams, or perhaps by an architect.
The individual pieces of the feature are then placed into the respective product backlogs of component teams (for example, the first piece is put into the component area 1 product backlog—“CA 1 PB” in the figure). Each component team performs Scrum against its own component-area-specific backlog, completing component specific pieces of end-customer features but not the full feature.
With two products the logistics of this problem are probably still manageable. However, what if the organization works on 10 or 15 products at the same time, and each of those products is dropping component-level pieces into the component team backlogs? At this scale the logistics of figuring out the proper order to work on the individual pieces within a particular component team backlog, while at the same time coordinating and integrating with all of the other component teams, become unmanageable.
Cross-Functional Feature Team
Cross-functional feature team have all of the skills necessary to work on multiple end-customer features and get them done—without having to farm out pieces to component teams. But what about the principal reason that most organizations create component teams—having a single trusted team to work in a component area? Won’t feature teams lead to chaotic development and maintenance of reusable components with large amounts of technical debt? Not if we have well-formed feature teams that, over time, share code ownership and collectively become trusted custodians of the code.
Multi-Feature Team Model
In this approach, the concept of a feature team has been reintroduced. There is now a single feature team that can pull an end-customer valuable feature off of the product backlog. This feature team has complete responsibility for doing the work and managing the logistics of getting the feature done.
Trusted component teams also remain in this model to help maintain the integrity of the individual component areas. These component teams still have a product backlog that typically contains technically oriented work that needs to take place within the component area (perhaps technical debt repayment work).
Also, as illustrated in Figure 12.3, a member of a component team can be assigned to be a member of a feature team. This person has the dual responsibility of being both a pollinator and a harvester. In the role of pollinator, the component team members pollinate feature teams with knowledge of the component areas to help better promote shared code ownership within the feature teams. In the role of harvester, component team members collect changes that the feature teams need to make within component areas and discuss those changes with their colleagues on the component teams, each of whom might also be collecting changes to the same component areas.
Scrum of Scrums
This practice allows multiple teams to coordinate their inter-team work. The team that performs the SoS is composed of individual members of the various development teams. Each development team determines which member to send to the scrum of scrums based on who can best speak to the inter-team dependency issues. Although I prefer to have consistency of representation, the person representing a team at the SoS can change over time based on who is best able to represent the team and speak to the issues at that point in time.
Participants at the scrum of scrums answer similar questions to the ones answered at the daily scrum:
- What has my team done since we last met that could affect other teams?
- What will my team do before we meet again that could affect other teams?
- What problems is my team having that it could use help from other teams to resolve?
Some teams timebox their scrum of scrums to be no more than 15 minutes, just like an individual Scrum team’s daily scrum. And they defer problem solving to occur after the scrum of scrums has completed so that only those participants whose involvement is necessary for problem resolution need attend.
A release train is an approach to aligning the vision, planning, and interdependencies of many teams by providing cross-team synchronization based on a common cadence. A release train focuses on fast, flexible flow at the level of a larger product.
The train metaphor is used to imply that there is a published schedule of when features will “leave the station.” All of the teams participating in the development of the product need to get their cargo onto the train at the appointed time. As in any country with reliable trains, the release train always departs on time and waits for no one. Likewise, if a team misses the train, it need not fret because there will be another train departing at a known time in the future. Leffingwell defines the rules of a release train as follows (Leffingwell 2011):
- Frequent, periodic planning and release (or potentially shippable increment—PSI) dates for the solution are fixed (dates are fixed, quality is fixed, scope is variable).
- Teams apply common iteration lengths.
- Intermediate, global, objective milestones are established.
- Continuous system integration is implemented at the top, or system, level, as well as at the feature and component levels.
- Release increments (PSIs) are available at regular (typically 60- to 120-day) intervals for customer preview, internal review, and system-level QA.
- System-level hardening iterations are (or may be) used to reduce technical debt and to provide time for specialty release-level validation and testing.
- For teams to build on top of similar constructs, certain infrastructure components—interfaces, system development kits, common installs and licensing utilities, user-experience frameworks, data and web services, and the like—must typically track ahead.
Also, as often as is practical, there should be system-wide integration and testing across the feature areas. Some teams reserve the last sprint before the train departs to be a time for hardening what has been developed in the previous sprints and integrating and testing the results across the various feature areas (for example, sprint 4 in Figure 12.5 might be a hardening sprint). As team skills mature, the need for a hardening sprint should diminish.