Rembrandt Tower, 14th floorAmstelplein 11096 HA Amsterdaminfo@sig.euwww.sig.eu

June 5, 2012

What Is the Value of Your Software?

3rd International Workshop on Managing Technical Debt - Zurich, 2012

Jelle de Groot, Ariadi Nugroho, Thomas Back and Joost Visser

I 13

Outline

BackgroundApproachExploratory StudyConclusion

2

Framing the notion of software valueI 13

Background

3

Cost to produce or replace

Produce

Transaction Price

Exchange Use

Future Benefits

Reliability: produce with stable quality

Conformance: to laws and

regulations

Optimization: more efficient

production

Innovation: new products or

servicesBased on actual software development expenses in

the past

Based on market average prices for software components and

development resources

Custom software assets are not commodities

traded on a competitive market

Value

Software value = Software production value

I 13

Background

Issues related to software value•Firms invest heavily on software, but often unsure of its value•Accounting practice tends to over estimate software value

4

How can we determine software value more objectively?Value

Technical debt as a component of software value•Technical quality affects software value•Poorly written code reduces value

I 13

Approach

5

Value = Area Value = Area - Debt

Debt

Time

Cost

Technical debt

Interest

Maintenance

Optimal maintenance

Impairment based on technical debt

I 13

Approach

Measuring technical debt•Based on SIG quality model of software maintainability•Quality gap to the “ideal” level determines debt•Debt estimate is based on more than 900 system snapshots

6

1-star 2-star 3-star 4-star 5-star

1-star

2-star 60%

3-star 100% 40%

4-star 135% 75% 35%

5-star 175% 115% 75% 40%

Rework Fraction (RF)SIG Quality Model

I 13

Approach

Proposed variants in measuring software value • V1: Subtracts repair cost (technical debt) from rebuild value• V2: Exclude parts that require rework from valuation• V3: Subtracts extra maintenance cost (interest) from rebuild value

7

V1: Fix the problem V3: Operate with higher costV2: Replace problematic parts

Applying the proposed valuation methods to 367 systems

ABAP COBOL C++ C# Java Mixed PL/SQL0

20

40

60

80

100

120

140

Frequency

Technology

I 13

Exploratory Study

8

Metrics Median

Size (KLOC) 77.0

Quality (star rating) 3.0

Rebuild value (MY) 7.8

Rework Fraction (%) 35.0

Repair effort (MY) 1.9

Descriptive Statistics

Comparison of three valuation modelsI 13

Explorative Study

9

0

2

4

6

8

10

Model 1 Model 2 Model 3

7.026.967.8

Average Software Value Across Models

Euro

s per

LOC The three valuation models

give similar values around €7 per LOC

Software value across different technologiesI 13

Explorative Study

10

C# and Java systems have the highest value. ABAP systems have the lowest value.

ABAP

0

5

10

15

20

Software Value per Line of Code - Model 1

COBOL C++ C# Java Mixed PL/SQLTechnology

EU

RO

S /

LOC

-5

0

5

10

15

ABAP COBOL C++ C# Java Mixed PL/SQLTechnology

EU

RO

S /

LOC

Software Value per Line of Code - Model 2

-5

0

5

10

15

Software Value per Line of Code - Model 3

EU

RO

S /

LOC

ABAP COBOL C++ C# Java Mixed PL/SQL

Technology

Technical debt across different technologies

I 13

Explorative Study

11

Java systems have the lowest technical debt averaging €1.6 per LOC

0

5

10

15

20

Technical Debt per Line of Code

ABAP COBOL C++ C# Java Mixed PL/SQLTechnology

EU

RO

S /

LOC

Conclusion•The notion of technical debt can be used to estimate the value of software that takes into account technical quality•No significant difference in the results given by the proposed valuation models (RE-based, RF-based, Interest-based)•C# systems have the highest value averaging €10 per LOC

Future Work•Compare results with traditional valuation approaches•Devise a method to estimate business value of software

I 13

Conclusion and Future Work

12

I 1313

I 13

Estimating Rework Fraction

Rework fraction is determined based on more than 900 system14

MonitorSnapshots S1Snapshots S2Snapshots S3Snapshots S3

Risk Profile

1-star2-star3-star4-star5-star

Moderate High Very High

31% 8% 4%16% 5% 4%13% 3% 0%10% 1% 0%7% 0% 0%

Rework Fraction

1-star2-star3-star4-star5-star

1-star 2-star 3-star 4-star 5-star

X 21% X 38% 17% X 43% 22% 5% X50% 29% 12% 7% X

[UI, US, UC, MC]

[UI, US, UC, MC]

1

2

3

For each quality leap:

RF = MAX(UI,US,UC,MC,Dup)

1-star 2-star 3-star 4-star 5-star

1-star

2-star 60%

3-star 100% 40%

4-star 135% 75% 35%

5-star 175% 115% 75% 40%

Rework Fraction (RF)