Top Navigation

The Scaled Agile Framework (SAFe) is the recognized approach for scaling Agile practices to the enterprise level. It provides comprehensive practices to help synchronize and coordinate Agile across many teams. However, companies adopting SAFe risk getting caught up in the many detailed elements of the framework and run a risk of losing the benefits of Agile. Companies must pay particular attention to accepted Agile principles as they use SAFe in order to keep SAFe agile.

Favor face-to-face collaboration over reliance on technology tools

SAFe helps coordinate activities across teams and technology tools provide support in this coordination. However, over reliance on technology tools can hinder face-to-face interaction. Frequent, personal communication speeds decision making and transfer of knowledge in a more complete way than through a tool. Visual display of team information through Program Boards, Sprint Backlogs, Kanbans, Product Backlogs, and other methods makes information more easily accessible than locked away inside of a tool.

Prioritize continuous integration to constantly realize value

It’s easy to fall into the mindset that the real delivery of the system happens at the end of the current Program Increment (PI). In reality, working code at the system level should be produced at the end of every iteration which provides the ability to ship new features every two weeks. In complex systems, this can be difficult. The Systems Team should be relentlessly working on how to provide automation everywhere in the delivery pipeline.

Cross-functional teams deliver fully working code

It is tempting to organize teams into familiar waterfall silos – business analysts, designers/developers, testers, and trainers. However, dividing teams creates boundaries which then requires handoffs introducing wait time and hindering efficient flow. Providing all the necessary skills to deliver an entire working piece of code reduces work in process and ensures value is delivered in the shortest amount of time. Focus teams on delivery.

Stable teams are the most productive

Waterfall project teams typically form and disband for each phase. In contrast, form individual teams and teams of teams within an Agile Release Train (ART) for stability. Learning as a collective team provides the best way to make the ART productive. Teams that have been together the longest tend to have the highest velocity. Team members should be dedicated full-time to avoid starting and stopping tasks as team members come and go.

Embrace changes

Don’t treat the Program Increment (PI) feature scope as fixed like in waterfall development. Keep some flexibility to adjust priorities during the PI so that important new features do not automatically get deferred 3 – 6 months in to the next PI. At the team level, features are typically locked down during the current Sprint (iteration), but new features can be added to the next iteration during the next Sprint Backlog Planning event. Work closely with teams throughout the PI to continually adjust feature priorities for possible inclusion in upcoming Sprints.

Business collaboration delivers the highest prioritized value

In waterfall development, the business typically sets the scope at the beginning of the project and accepts the solution at the end of development. Don’t let Program Increments fall into this same pattern where business participation occurs at the beginning in PI Planning and at the end for the Solution Demo. Keep the Business Owners and Product Management/Owners closely involved with the teams throughout the entire Program Increment so that features can be reprioritized throughout development to deliver the highest value.

Keep it simple

Small batch sizes increase velocity. Keep the number of features to be worked on in the current Program Increment small. The waterfall approach overloads as many features as possible into the current phase for fear that there is only one chance to request features. Don’t allow PI planning to pack in too many features. With continual prioritization of the backlog and constant two week cadence of iterations, features can be broken down into small, high-value increments with high confidence of on-time delivery.