CS207 #3, 8 Oct 2010Gio Wiederhold

http://infolab.stanford.edu/people/gio.htmlGates B12

Apr 20, 2023 1CS207

Syllabus:1. Why should software be valued? 2. Open source software. Scope. Theory and reality3. Principles of valuation. Cost versus value. 4. Market value of software companies. 5. Intellectual capital and property (IP). 6. Life and lag of software innovation. 7. Sales expectations and discounting. 8. The role of patents, copyrights, and trade secrets. 9. Alternate business models. 10. Licensing. 11. Separation of use rights from the property itself. 12. Risks when outsourcing and offshoring development. 13. Effects of using taxhavens to house IP. Apr 20, 2023 2CS207

Review definitions: Intangibles

• Software is an intangible goodIf it is owned it is considered Intangible Property

In a business there are 3 parts that have value(Contribute to potential income)

1. Tangible goods: buildings, computers, money2. The know-how of management & employees3. Intellectual property: Software, patents, etc.

2. + 3. make up the Intangible Capital of a company.

3Apr 20, 2023 CS207

4

Basis for SW value as of today

• Sum of future income Sales = price * copy count Maintenance fees if service subscription

• Minus sum of future costs Cost of goods Cost of marketing Cost of doing business Cost of maintenance

• Discounted to today To account for risk Inde

pend

ent o

f cos

t

Apr 20, 2023 CS207

Direct Valuation

• Value the software specifically by income over its lifetime

• But software is not stable over time: SlitheryGetting long-term income requires maintenanceMaintenance enables long-term income

• Much more so than other intangibles Books, music,

• Similar to some intangibles that contribute to life Costumer loyalty, trademarks

04/20/23 CS207 Fall 2010 5

Maintenance is beneficial

Lif

etim

e m

ain

ten

ance

co

std

epre

ciat

ion

/ y

ear

= 1

/ l

ifet

ime

100%100%

4040

00

2020

7070

3030

1010

8080

9090

6060

5050

life

04/20/23 6CS207 Fall 2010

yearsyears

44

22

77

33

11

88

99

66

55

1313

11111212

1010

PCs cars software intangiblesPCs cars software intangiblesTypical Life Typical Life 3years 3years 5 years 5 years 112 years 18 years2 years 18 yearsMaintenance 2%/year 5%/year 15%/year 13.75%/year Maintenance 2%/year 5%/year 15%/year 13.75%/year Maintenance cost 6%Maintenance cost 6% 21%21% 80%80% most most over asset life over asset life Depreciation Depreciation 33/y. linear33/y. linear 20%/ y. linear20%/ y. linear 8%/y. linear8%/y. linear 12% geometric12% geometric

7

Software is slithery !

Continuously updated1. Corrective maintenance

bugfixing reduces for good SW

2. Adaptive maintenance externally mandated

3. Perfective maintenancesatisfy customers' growing

expectations[IEEE definitions]

Life time

Ratios differ in various settings

100%

80%

60%

40%

20%

04/20/23 CS207 Fall 2010

8

IP sources• Corrective maintenance

Feedback through error reporting mechanisms Inadequate protection from virus etc. Taking care of missed cases Complete inadequate tables and dimensions

• Adaptive maintenanceStaff to monitor externally imposed changes

Compliance with new standards Technological advances

• Perfective maintenanceFeedback through sales & marketing staff

Minor features that cannot be charged for04/20/23 CS207 Fall 2010

9

Technical Parameters needed

IP is to be valued as of some specific date

1. Life of the IP in the product from that time on

The interval from completion until little of the original stuff is left

2. Diminution of the IP over the Life

A bit like a depreciation schedule, but based on content replacement, until little IP is left. 10% is a reasonable limit.

3. Lag*, interval from transfer to start of IP diminution

= the time before an investment earns revenue• also called “Gestation Period

4. Relative allocation, if there are multiple products contributing to income.

design,code, . . . .

Apr 20, 2023 CS207

Crucial assumption

• IP content is proportional to SW size Not the value, that depends on the income=======================================

Pro: Programmers efforts create code An efficient organization will spend money wisely

Counter: not all code contributes equally early code defines the product, is most valuable new versions are purchased because of new features

• Arguments balance out it is the best metric we can obtain

Apr 20, 2023 CS207 10

11

Maintenance → SW Growth

Rules: Sn+1 = 2 to 1.5 × Sn per year [HennesseyP:90]

Vn+1 ≤ 1.30% × Vn [Bernstein:03]

Vn+1 = Vn + V1 [Roux:97] ([BeladyL72], [Tamai:92,02] indications) [Blum:98] [Blum98] [Blum:98]Deletion of prior code = 5% per year [W:04]

at 1.5 year / version

Apr 20, 2023 CS207

12

Observations

• Linear growth has been observed, is reasonable• Software cannot grow exponentially

Because no Moore's Law 1. Cost of maintaining software grows exponentially with size

The number of interactions among code segments grow faster [Brooks:95]

2. Can't afford to hire staff at exponential *2

3. Cannot have large fraction of changes in a version And get it to be reliable

4. Cannot impose version changes on users < 1 / year5. Deleting code is risky and of little benefit

except in game / embedded codeApr 20, 2023 CS207

13

Price remember IP = f(income)

• But --- Price stays ≈ fixed over timelike hardware Moore's Law

Because1. Customers expect to pay same for same functionality2. Keep new competitors out 3. Enterprise contracts are set at 15% of base price4. Shrink-wrapped versions can be skipped

• Effect

The income per unit of code reduces by 1 / size →Apr 20, 2023 CS207

14

Growth diminishes IP

at 1.5 year / version

For constant unit price

Apr 20, 2023 CS207

15

Total income

Total income = price × volume (year of life)

• Hence must estimate volume, lifetimeBest predictors are Previous comparables

Erlang curve fitting (m=6 to 20, 12 is typical)

and apply common sense limit = Penetration estimate total possible sales F × #customers above F= 50% monopolistic aberration

P

Apr 20, 2023 CS207

16

Sales models1. Normal curve: simple, no defined start point2. Erlang: realistic, more complexboth have same parameters: mean and variance

Apr 20, 2023 CS207

Growth and Perception

E-commerce [this slide based on a 2001 CS99/73N class exercise]

• Gartner: 2000 prediction for 2004: 7.3 T$• Revision:2001 prediction for 2004: 5.9 T$ drastic loss?

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 ...

Realistic

growth

Perceived growth

Invisible growth

Extrapolated growth

Disap- pointment Combi-

natorial growth

Perceived initial growthPerception level

ExamplesArtificial IntelligenceDatabasesNeural networksE-commerce

50 companies, each after 20% of the market

Failures

Apr 20, 2023 17CS207

T r e n d s 1998 : 1999

• Users of the Internet 40% 52% of U.S. population• Growth of Net Sites (now 2.2M public sites with 288M pages)• Expected growth in E-commerce by Internet users [BW, 6 Sep.1999]

segment 1998 1999 books 7.2% 16.0% music & video 6.3% 16.4% T o y s 3.1% 10.3% travel 2.6% 4.0% tickets 1.4% 4.2% Overall 8.0% 33.0% = $9.5Billion

An unsustainable trend cannot be sustained [Herbert Stein, Council Econ. Adv, 1974]

new services

98 99 00 01 02 03 04 0.3 1 3 9 27 81 **

90 80 70 60 50 40 30 20 10

0

Year / %

%

Centroid, in 1999 ~1% of total market

E-penetration Toys

Apr 20, 2023 18CS207

19

%

100

90

80

70

60

50

40

30

20

10

00 1 2 3 4 5 years

Vn Vn+1 Vn+2

DepreciationNormal Erlang or Weibull

Sales curves

Apr 20, 2023 CS207

20

Erlang sales curves m=mean/variance

0

2,000

4,000

6,000

8,000

10,000

12,000

14,000

16,000

18,000

Erlang m = 12

Erlang m = 6

| | end of time horizon| 9 years|

^ 50,000 w hen| Erlang m ~ infinite

For 50 000 units over 9 years

Flash-in-the pan

One-time promotion

Long-lived single product

Apr 20, 2023 CS207

21

Ongoing Version Sales

Product Line sales

-

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17

years

sa

les

ReplacementProduct

approximation

Predicted product sales for 5 versions, stable rate of product sales3 year inter-version interval, first-to-last product 12 years, life ~15 years

Apr 20, 2023 CS207

Apr 20, 2023 CS207 22

Fraction of income for SW

Income in a software company is used for

• Cost of capital typicalDividends and interest ≈ 5%

• Routine operations -- not requiring IP Distribution, administration, management ≈ 45%

• IP Generating Expenses (IGE)Research and development, i.e., SW ≈ 25%Advertising and marketing ≈ 25%

These numbers are available in annual reports or 10Ks

Apr 20, 2023 CS207 23

Recall: Discounting to NPV

Standard business procedure• Net present Value (NPV) of

getting funds 1 year later = F×(1 – discount %)Standard values are available for many businesses

based on risk (β) of business, typical 15%Discounting strongly reduces effect of the far future

NPV of $1.- in 9 years at 15% is $0.28

Also means that bad long-term assumptions have less effect

24

Example

Software product Sells for $500/copy Market size 200 000 Market penetration 25% Expected sales 50 000 units Expected income $500 x 50 000 = $25M What is the result?

Apr 20, 2023 CS207

25

Combining it allfactor today y1 y2 y3 y4 y5 y6 y7 y8 y9

Version 1.0 2.0 3.0 4.0 5.0 6.0 7.0

unit price $500 500 500 500 500 500 500 500 500 500

Rel.size 1.00 1.67 2.33 3.00 3.67 4.33 5.00 5.67 6.33 7.00

New grth 0.00 0.67 1.33 2.00 2.67 3.33 4.00 4.67 5.33 6.00

replaced 0.00 0.05 0.08 0.12 0.15 0.18 0.22 0.25 0.28 0.32

old left 1.00 0.95 0.92 0.88 0.85 0.82 0.78 0.75 0.72 0.68

Fraction 100% 57% 39% 29% 23% 19% 16% 13% 11% 10%

Annual $K 0 1911 7569 11306 11395 8644 2646 1370 1241 503

Rev, $K 0 956 3785 5652 5698 4322 2646 1370 621 252

SW IP 25% 0 239 946 1413 1424 1081 661 343 155 63

Due old 0 136 371 416 320 204 104 45 18 6

Disct 15% 1.00 0.87 0.76 0.66 0.57 0.50 0.43 0.38 0.33 0.28

Contribute 0 118 281 274 189 101 45 17 6 2

Total 1 032 ≈ $ 1 million Apr 20, 2023 CS207

Apr 20, 2023 CS207 26

Result of Example

• Selling 50 000 SW units at $500 ≈ $ 1M not $ 25M

Once its in a spreadsheet, the effect of the many assumptions made can be checked.

When assumptions later prove unwarranted then management can make corrections.

To be wise, don't spend more than ≈ $500 000 to develop the software product.

Guidance obtained

• We applied an overall Erlang sales curve new versions keep market going but customers

do not replace earlier versions

• The assumption are sufficiently simple that alternatives can be intelligently discussed1. keep development costs low2. design so that SW maintenance is low3. charge a higher price4. broaden the market5. or →

Apr 20, 2023 CS207 27

Business models0. New versions do not replace earlier versionsAlternative business models 1.New versions encourage replacement 2.Provide related services3.Charge for maintenance

Lower initial cost, slower income stream

4.Make product Open source to broaden marketCharge only for services

Apr 20, 2023 CS207 28

Discussion• Many choices nowa. Technicalb.BusinessInteract with each other.

Apr 20, 2023 CS207 29