MTAT Software Economics. Workshop 7: Software Cost Estimation (part II)

Transcription

1 MTAT Software Economics Workshop 7: Software Cost Estimation (part II) Marlon Dumas marlon.dumas ät ut. ee

2 Outline Presentations of Effort Estimation Homework Discussion on book chapter by Caper Jones Agile Estimation Earned Value Analysis 2

3 Six forms of software cost estimation (by Caper Jones) Project Phase Activity Project Phase Activity 3

4 Discussion (cont.) What is the difference between a phase, an activity and a task? How many phases/activities/ tasks are typical in a software project? What key issue arises when estimating only at the level of project/phases? Does it mean that projectlevel estimation is useless? According to the study reported in the chapter, which form of cost estimation is most accurate? In which context is manual estimation most used? In which context is automated estimation used? 4

5 Agile Estimation See Mike Cohn s slidedeck 5

6 Project Monitoring using Earned Value Analysis

7 Project Monitoring Cost estimation is about estimating how much will be spent by a project/phase/activity. Cost estimation is a planning tool Project monitoring is about assessing how much has been done in order to answer the question: Are we on track to deliver on-time and on budget? Earned Value Analysis (EVA) 7

8 Earned Value Analysis (EVA) Technique for comparing actual project progress against planned progress Rooted in early 20 th century industrial manufacturing Introduced in the 1960s to monitor United States Department of Defense projects Nowadays considered as a project management technique (Earned Value Management) Standardized: ANSI/EIA 748-A standard 8

9 For discussion What do you think about the following reasoning: We re mid-way through the project timeline and we ve only implemented 30% of the features. We re late?? We re mid-way through the project timeline, we ve spent 50% of the budget and 50% of the work is done. So we re doing well?? 9

10 Some Terminology WBS (Work Breakdown Structure) Decomposition of a project into phases/activities/tasks We consider WBS where each unit is annotated with an estimate of duration and cost Can be obtained from the cost estimation Planned value at time t How much will have been spent at time t according to the plan? 10

11 Project Monitoring: A Naïve Approach Source: Wikipedia 11

12 Let s get a bit more sophisticated Three basic measures: Budgeted Cost of Work Performed (BCWP) also called Earned Value Budgeted Cost of Work Scheduled (BCWS) also called Planned Value Actual Cost of Work Performed (ACWP) also called Actual Cost 12

13 The three measures in a nutshell Source: Wikipedia 13

14 And a bit more sophisticated Two derived measures: SV: Schedule Variance (BCWP-BCWS) CV: Cost Variance (BCWP-ACWP) 14

15 Schedule Variance Source: Wikipedia 15

16 Cost Variance Source: Wikipedia 16

17 A bit more Three more derived measures: SPI: Schedule Performance Index = BCWP/BCWS SPI < 1 project is behind schedule CPI: Cost Performance Index = BCWP / ACWP CPI < 1 project is over budget CSI: Cost Schedule Index = CPI x SPI The further CSI is from 1.0, the less likely project recovery is Question: What if SPI or CPI are very high? 17

18 Forecasting BAC Budget at Completion IEAC (Independent) Estimate At Completion = BAC / CPI ISAC Independent Schedule At Completion = Schedule / SPI (Schedule is the planned project duration) 18

19 Let s get our hands dirty Quick exercise: 19

20 EVA after the deadline If the project has already passed its planned deadline, BCWP will converge to BCWS and SPI becomes useless since it tends to 1 A more appropriate measure of delay in this case is: Days Ahead: how many days ahead of schedule we are (negative if behind schedule), calculated from the original schedules of the individual tasks. 20

21 Pros: Pros and cons of EVA Provides early warning of cost/schedule problems Enables forecasting final project cost/schedule Allows us to trace impact of key events on a project Cons: High data collection effort May be distorted by inaccurate work-in-progress measures (some authors recommend to grade work using only three values: 0%, 50% or 100% completed Significant effort required if the requirements/scope change significantly. 21

22 Agile Project Monitoring In agile methods, relatively little upfront effort is spent in requirements analysis and planning Requirements creep is intentionally high WBS is very dynamic EVA may be inappropriate as primary monitoring tool (although it can still be used externally, outside the agile team) 22

23 Agile Project Monitoring Measure of effort: story point (or ideal day) Project decomposition unit: release/iteration Productivity measure: Velocity (e.g. story points per iteration) Visual tracking techniques Burndown chart (curve) Burndown bar chart takes into account work completed and scope change 23

24 Burndown Chart Burn-up event Story points 24

25 Burndown Bar Chart Story points Work Added Work Remaining

26 Forecasting Completion from Burndown 26

27 Homework (5 points) Read the following tutorial Complete the exercise in the Appendix of the above tutorial (pages 12-14). Submit the homework (3 pages) either via: The following submission page (you ll need to create a user account with Orangut): Or give it to the lecturer on paper next Tuesday Deadline: Tuesday 19 October at 11:59pm GMT One submission per person 27

28 Acknowledgments Some material is adapted from: asguest evaslides-education-pptpowerpoint/ (earned value analysis) Wikipedia article on Earned Value Management (burndown charts) 28