1. CS 5150 Software Engineering Lecture 6 Project Management

2. CS 5150 2 Administrivia Quizzes posted Feasibility study/plan due in a little over a week

3. CS 5150 3 Classic Project Management Example: OS 360 The OS for the IBM 360 was 2 years late Q: How does a project get 2 years behind schedule? A: One day at a time! Fred Brooks Jr, The Mythical Man Month

4. CS 5150 4 “The Mythical Man Month” Adding more people to work on a task helps up to a point, then becomes counter- productive Adding people adds coordination overhead Some activities work best in a single mind Some tasks cannot be accelerated beyond some limit In software, productivity varies A LOT between developers Some studies found that the most productive are 5x faster than the “just adequate”

5. CS 5150 5 The Goal of Project Management Give the team the best chance to finish a project on time on budget etc Provide visibility into how well things are going according to plan

6. CS 5150 6 The Central Tension in Project Management Clients (or managers, or whoever) want to know: What will the functionality be? When will it be available? What will it cost? Examples: For products, marketing needs to how to sell the software For embedded systems, other teams need to coordinate development

7. CS 5150 7 The Central Tension in Project Management BUT: Every software project is a special snowflake Time and cost estimates are notoriously inaccurate Almost always something changes about the specifications during development

8. CS 5150 8 Typical Project Planning Experience Some big projects (e.g. product releases) some smaller project (e.g. “phase 1” contracts) For smaller projects, just create rough time estimates and judge whether they’re reasonable For larger projects, detailed planning with labor estimates and task dependencies Still usually off by quite a bit Unanticipated tasks Unanticipated work on planned tasks Unanticipated schedule conflicts Unanticipated life conflicts

9. CS 5150 9 Conventional Software Project Management The scope of the project is defined early Development is divided into tasks and milestones Time and resource estimates are made Estimates are combined to create a schedule/plan Progress relative to the plan is monitored, perhaps weekly The plan is modified as inaccuracies in the estimates become evident

10. CS 5150 10 Agile Project Management Planning is divided into high level release forecasting and low level detail planning Release planning is a best guess of what can be achieved in a sequence of time-boxes (sprints) For each time-box, small teams do detailed planning Release plan is updated to reflect sprint results Clients and developers share ownership of the release plan and choice of sprints

11. CS 5150 11 Universal Aspects of Project Management Planning 5150: Rough plan for feasibility study More detailed plan for each milestone Progress tracking Regular comparison of plan and reality Update the plan! Occasional reassessment of project management processes

12. CS 5150 12 Generating Time Estimates Developers are usually pretty good at estimating times for tasks within their areas of expertise, but... Even so, little things get in the way The difference between “almost done” and “completely done” Unplanned technology churn There is always something that is a little bit different

13. CS 5150 13 Team-based Estimation Developers themselves usually produce the best estimates within a small time window Teams can commit to the outcome of a sprint Still useful to have a local schedule/plan Different teams work at dramatically different rates Reminder: 5150 is about 1 sprint’s worth of development time

14. CS 5150 14 Start-up Time Recruiting, tying up loose ends from previous projects Acquiring and setting up equipment Learning about new domains and frameworks Getting the ball rolling with clients/other collaborators Can be months for large projects

15. CS 5150 15 Methods for Task-Based Project Planning Critical path method, Gantt charts, etc Build a work plan from task data/estimates Display plan in graphical and tabular forms Software support Spreadsheets/databases for recording estimates and results Fancier software for calculating estimates and generating reports

16. CS 5150 16 Task-Based Project Planning Terminology Task/activity A part of a project that takes work/time Event The completion of a group of activities Dependency Constraints imposed on one activity by another Resource Staff time, equipment or other item required for a task

17. CS 5150 17 Task-Based Project Planning Terminology Deliverable Any work product that is provided to clients or users (software, documentation, etc) Milestone A set of tasks, usually associated with a target date

18. CS 5150 18 A Simple Gantt Chart (made with Microsoft Excel)

19. CS 5150 19 Gantt Charts Most useful for small projects or pieces of larger projects Horizontal axis is time Each row is a task: (planned) start time, portion complete, (projected) completion time Progress can be tracked by drawing a vertical line at the current date

20. CS 5150 20 A More Complex Gantt Chart (Made with SmartDraw)

21. CS 5150 21 Alternative Visualization: Activity Graph

22. CS 5150 22 Activity Graph Background PERT Program Evaluation and Review Technique introduced by the U.S. Navy in 1957 to support the development of its Polaris submarine missile program. PERT/Time Activity graph with three time estimates (shortest, most probable, longest) on each activity to compute schedules. PERT/Cost Added scheduling of resources (e.g., facilities, skilled people, etc.)

23. CS 5150 23 Critical Path Method Uses and activity graph with a single time estimate for each task A standard method for managing large construction projects On big projects, activity graphs with 10,000+ tasks are common

24. CS 5150 24 Step 1: Time Estimates

25. CS 5150 25 Step 2: Earliest Dates

26. CS 5150 26 Step 3: Latest Dates

27. CS 5150 27 Step 4: Slack

28. CS 5150 28 Step 5: Critical Path

29. CS 5150 29 Feedback One of the latest trends in project management: evidence-based scheduling Developers generate estimates for all their tasks and record the actual time taken A learning algorithm adapts to each developer’s estimation strengths and weaknesses over time