Chapter 23 Estimation Software Engineering: A Practitioner’s Approach 6 th Edition Roger S. Pressman

2 Software Project Estimation (1) S/W is the most expensive element of virtually all computer based systems S/W cost and effort estimation will never be an exact science –Too many variables Human Technical Environmental Political

3 Software Project Estimation (2) Options for estimation –Delay estimation until late in the project Attractive, but not practical –Base estimates on similar projects that have already been completed Unfortunately, past experience has not always been a good indicator of future results –Use relatively simple decomposition techniques to generate project cost and effort estimates “Divide and conquer” approach –Use one or more empirical models for software cost and effort estimation Can be used as a cross-check for the previous option and vice versa

4 Decomposition Techniques Two different points of view for the decomposition approach –Decomposition of the problem –Decomposition of the process But first, the project planner must –Understand the scope of the s/w to be built –Generate an estimate of its “size”

5 Software Sizing (1) The accuracy of a s/w project estimate is predicated on a number of things: –The degree to which the planner has properly estimated the size of the product to be built –The ability to translate the size estimate into human effort, calendar time, and dollars (required availability of past records) –The degree to which the project plan reflects the abilities of the s/w team –The stability of product requirements and the environment that supports the s/w engineering effort

6 Software Sizing (2) Sizing represents the project planner’s first major challenge Size refers to a quantifiable outcome of the s/w project (e.g. LOC and/or FP) Four different approaches to the sizing problem [PUT92] –“Fuzzy Logic” sizing –Function point sizing –Standard component sizing –Change sizing

7 Problem-Based Estimation (1) Example of baseline productivity metrics are LOC/pm or FP/pm Making the use of single baseline productivity metric is discouraged In general, LOC/pm or FP/pm averages should be computed by project domain Local domain averages should be used

8 Problem-Based Estimation (2) Statistical approach – three-point or expected-value estimate S = (s opt + 4s m + s pess )/6 –S = expected-value for the estimation variable (size) –s opt = optimistic value –s m = most likely value –s pess = pessimistic value

9 An Example of LOC-Based Estimation (1) FunctionEstimated LOC User interface and control facilities (UICF) Two-dimensional geometric analysis (2DGA) Three-dimensional geometric analysis (3DGA) Database management (DBM) Computer graphics display facilities (CGDF) Peripheral control function (PCF) Design analysis modules (DAM) 2,300 5,300 6,800 3,350 4,950 2,100 8,400 Estimated lines of code33,200

10 An Example of LOC-Based Estimation (2) Estimated lines of code = W = 33,200 Let, –Average productivity = 620 LOC/pm = X –Labor rate = $8,000 per month = Y So, –Cost per line of code = Z = Y/X = $13 (approx.) –Total estimated project cost = W*Z = $431,000 (approx.) –Estimated effort = W/X = 54 person-months (approx)

11 An Example of FP-Based Estimation (1) Information Domain Value CountWeighting factor SimpleAverageComplex External Inputs (EIS)3X346=9 External Outputs (EOs)2X457=8 External Inquiries (EQs)2X346=6 Internal Logical Files (ILFs) 1X71015=7 External Interface Files (EIFs) 4X5710=20 Count Total50 Figure 15.4: Computing function points

12 An Example of FP-Based Estimation (2)

13 An Example of FP-Based Estimation (3) Value Adjustment Factors

14 An Example of FP-Based Estimation (4) Now, –FP estimated = count-total  [   (F i ) ] F i (i = 1 to 14 are value adjustment factors) So, –FP estimated = W = 320  [  52] = 375 (approx.) Let, –Average Productivity = X = 6.5 FP/pm –Labor rate = Y = $8,000 per month So, –Cost per FP = Z = Y/X = $1,230 (approx.) –Total estimated project cost = W*Z = $461,000 (approx.) –Estimated effort = W/X = 58 person-months (approx)

15 Empirical Estimation Models Uses empirically derived formulas to predict effort as a function of LOC or FP The empirical data are derived from a limited sample of projects So, no estimation model is appropriate for all classes of s/w and in all development environments The results obtained from such models must be used judiciously An estimation model should be calibrated to reflect local conditions

16 The Structure of Estimation Models (1) Derived using regression analysis on data collected from past s/w projects Overall structure, E = A + B  (e v ) c Here, –A, B, and C are empirically derived constants –E is effort in person-months –e v is the estimation variable (either LOC or FP) Some form of project adjustment component is also used

17 The Structure of Estimation Models (2) Example of a LOC-oriented estimation model (Bailey-Basili model)  E =  (KLOC) 1.16 Example of a FP-oriented estimation model (Kemerer model)  E = FP Estimation models must be calibrated for local needs.

Summary of S/W estimation methods 18 Usecase based Problem based Process based LOC FP Man- month E(LOC) LOC/usecase E(FP) E(Effort) COST $/FP $/LOC $/Man-month TIME S/W equation

19 The COCOMO II Model (1) COnstructive COst Model A hierarchy of estimation models Addresses the following areas –Application composition model –Early design stage model –Post-architecture stage model Three different sizing options are available –Object points –Function points –Lines of source code

20 The COCOMO II Model (2) The COCOMO II application composition model uses object points Object point is computed using counts of the number of –Screens (at the user interface) –Reports –Components likely to be required to build the application

21 The COCOMO II Model (3) Object type Complexity weight SimpleMediumDifficult Screen123 Report258 3GL component10 Figure 23.6: Complexity weighting for object types

22 The COCOMO II Model (4) Developer’s experience/capability Very low LowNominalHigh Very High Environment maturity/capability Very low LowNominalHigh Very High PROD Figure 23.7: Productivity rate for object points

23 The COCOMO II Model (5) NOP = (object points)  [(100-%reuse)/100] –NOP = New Object Points –Object Points = Weighted Total –%reuse = Percent of reuse Estimated effort = NOP/PROD –PROD = Productivity Rate –PROD = NOP/person-month

24 The Software Equation [PUT92] (1) It is a multivariate model Has been derived from productivity data collected for over 4000 contemporary s/w projects E = [LOC  B /P] 3  (1/t 4 ) Here, –E = effort in person-months or person-years –t = project duration in months or years –B = “special skills factor” –P = “productivity parameter”

25 The Software Equation [PUT92] (2) For small programs (KLOC = 5 to 15) –B = 0.16 For programs greater than 70 KLOC –B = 0.39 P = 2000 for development of real-time embedded s/w P = 10,000 for telecommunication and systems s/w P = 28,000 for business systems applications

26 The Software Equation [PUT92] (3) Simplified formulas –t min = 8.14 (LOC/P) 0.43 in months for t min > 6 months t min = minimum development time –E = 180 B t 3 in person-months for E  20 person-months Here t is represented in years

27 The Make/Buy Decision S/W acquisition options –s/w may be purchased (or licensed) off the shelf –“full-experience” or “partial-experience” s/w components may be acquired and then modified and integrated to meet specific needs –s/w may be custom built by an outside contractor to meet the purchaser’s specifications

28 Creating a Decision Tree (1)

29 Creating a Decision Tree (2) The expected value for cost, computed along any branch of the decision tree, is –Expected cost =  (path probability) i  (estimated path cost) i –Where, i is the decision tree path

30 Outsourcing S/W engineering activities are contracted to a third party who does the work at lower cost and, hopefully, higher quality S/W work conducted within a company is reduced to a contract management activity The decision to outsource can be either strategic or tactical Has both merits and demerits

31 Chapter 23 Introduction 23.5 to Exercises –23.1, 23.4—23.8, 23.12