cs207 #3, 8 oct 2010 gio wiederhold gates b12 23-oct-151cs207
TRANSCRIPT
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