When portfolio and program managers undertake quarterly (or annual) feature and portfolio planning, understanding team dependencies is made necessary to identify constraints and avoid overburdening one team. Complexity quickly grows beyond intuition after two or three team dependencies are identified, rendering current forecasting techniques unsatisfactory. This session will discuss the inadequacies of current tools and frameworks used to manage the dependency planning problem, and introduce the choices you have for reducing the complexity and new planning techniques that untangle dependencies into a doable plan.
At the end of the session, attendees will have a new understanding of the complexity team dependencies add to planning, and have a set of strategies and techniques to predictably manage products built in high team-dependency organizations. This session looks at example dependency graphs and graphical matrix techniques that are quick to build and give clear risk insight.
It often shocks organizations to learn mathematically that each team dependency HALVES the chances of an on-time completion of component or delivery. With two dependencies, there is a 1 in 4 chance of no delay; with three dependencies, there is a chance of 1 in 8 delivering on-time. One large legacy application the author worked with had seven dependencies from a core library to a user interface – that is a 1 in 128 chance no team will be delayed (127 times more likely to experience one or more delays). Planning clearly needs to consider how dependencies might impact each teams ability to integrate and build.
It is not as dire as it sounds, not every team suffers the same chance of delay. We look at how to analyze historical examples of delayed work to identify types of features that will encounter dependency delays in the future. Building a map (linked graph) and matrix visualizations of team dependencies gives a basis for examining this historical likelihood of delay and planning team organization structures or staffing plans that compensate. It is possible to predictibly plan in high dependency environments, its just too hard to do in ad-hoc ways.
