Software development and making money from it:

1. CS207 #5, 28 Oct. 2011
Gio Wiederhold
Gates B12
10/28/2011 Gio: CS207 1

2. Syllabus:
1. Why should software be valued?
2. Open source software. Scope. Theory and reality
3. Principles of valuation. Cost versus value.
4. Market value of software companies.
5. Alternate business models. Making money from free SW
6. Intellectual capital and property (IP).
7. Life and lag of software innovation. Marketing. Allocation.
8. Sales expectations and discounting.
9. The role of patents, copyrights, and trade secrets.
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.
10/28/2011 Gio: CS207 2

3. Example of Free
• Adobe produced software to generate and read
markup text (pdf) for sale to companies.
 minor business for internal publishing
• Arrangement with the IRS that if Adobe would
separate the reader and provide for free, it would
publish tax forms using pdf
 huge business – now everyone needed a reader and
companies bought pdf generators to publish in pdf
• When patents ran out, others companies made pdf
generators available
 Adobe still provides many pdf related services
10/28/2011 Gio: CS207 3

4. Let’s ignore the
intangibles, we
cannot measure
them reliably.
Book
value
Intangibles
10/28/2011 Gio: CS207 4

5. Timing of expense
and income
capitalization of cost
Manufacturing &
allowed under GAAP distribution delay
Release to Production
←→ Distribution
to Sales
Sales
← ─→
Revenues →
development lag Centroid of
Costs →
Centroid of total revenue
development cost .. Testing
~60%→
Sales lag .
Research,
Design, Implementation
← −−−−−−−−−−−−− →
Development done → ←→ marketing lag time →
Marketing
Centroid of pre-
← Costs
sales marketing costs Post-sales marketing,
part of investment: IGE part of sales cost: CoGS
28-Oct-11 CS207 5

6. SW Lags
Development done → β ← General
Revenues →
availability
↔ Sales
development lag ←→
Costs →
includes testing Centroid of revenue
&Test
ing ←−−−−−−−→ sales lag .
Research,
Implementation
← Costs
Centroid
of pre-sales ↔ marketing lag time →
marketing costs Marketing
28-Oct-11 CS207 6

7. Lag delays benefits
of R&D investments
Estimate
effective lag .
growth limit
~37% →
Effort →
@27.4% →
~14% →
Testing
growth limit Development
Research
35%→
start 75% 50% 25% done
10/28/2011 Gestation period → Gio: CS207 7

8. Start-up
development
A startup is unlikely to ramp up linearly
Use exponentional growth, exp 0.025
Assume
1. 12.5% research
Given that idea is clear, only towards for implementation
2. 25.0% testing
Minimal and risky
3. 67.5% left for implementation
• Overlap research and implementation until testing starts
• Overlap implemtation and testing until RPS
Results Overall centroid @0.27 before RPS -- later
Research from 1.00 to 0.33, centroid @ 0.65 before RPS
Implementation from 0.67 to 0.00, centroid @ 0.29 before RPS
Testing from 0.17 to 0.00, centroid @ 0.08 before
Hiring rate at RPS 21%, at the limit for effectiveness Ignore different
28-Oct-11 CS207 staff salaries 8

9. Graph of start-up
development
21% effort
Testing starts when
growth
17% time remains →
100% Res., Imp, &
Implementation starts when Test @0.27 → 25%
80%
67% time remains → Testing
60% Research ends when
33% time remains →
40%
12.5% Research 62.5%
20% @0.65→ Implementation
@0.29→
0
start 0.75 0.50 0.25 done
28-Oct-11 CS207 9

10. Development in mature
company with
12.5% research and
25% testing effort,
38% effort growth 62.5% implementation
at start Res., Imp, & 5% company staff growth
Test @0.42 →
100%
Available Testing starts when
Relative Effort →
←−−− 40% time remains
75% resources Research ends when
← 65% time remains T
←Implementation starts when
50% 85% time remains
R
I
25% @0.46→
0 @0.85→
1.00 0.75 0.50 0.25 done
Values based on finite integration, exp= 0.05
28-Oct-11 CS207 10

11. Lag differs less than
development period
Testing
R&I done
done
start
Testing
R&I
start done
Testing
Effective lag =
Development period × Centroid fraction R&I
done
start done
28-Oct-11 CS207 11

12. Ongoing
development
New considerations
1. Have staff already
a. Early versions rapid growth, but observe ~20% limit
b. Later, best grow slower
2. Can overlap version development
a. Don’t let valuable staff be idle
b. Missing features should already be understood
c. Rapid analysis of problems to allow next version fixes
d. Any research should be done before major staff effort
3. Adequate testing to keep reputation
CS207 12

13. 2nd Version
substantial testing
100%
Effort →
43% Testing
during version n
56% 75% All@0.61→ interval
T@0.33→
50%
R&I @1.00 →
25%
Release
1.50 1.25 version n-1 0.75 0.50 0.25 done
Staff becomes available when prior version enters testing
28-Oct-11 CS207 13

14. Marketing
• Business model must allocate spending optimally
 Technology, as needed, long life and lag
&
 Marketing, necessary, less lag, slower growth
 Life of advertising 50% of technology, mix product & brand
• Interdependence viral
in your
brain
forever
 Consistent
 Relevant CS207
 Linked by a common name and label
 Honest name for file software misled: FLASH for flexible, but it wasn’t fast
28-Oct-11 CS207 14

15. Planning:
Consistency in plans
When comparing business alternatives
• Give each choice the same chance
1. Temporal consistency
 Computing versus communication
 Local versus Cloud in 2012
o Skate to where the puck is going [Gretsky]
2. Discount rate
3. Resource prices
 Green alternatives
 Benefits may depend on future price of oil –
o if you assume future price = 3 x now, why not invest in oil instead
10/28/2011 CS207 15

16. Example
Enterprise SW versus cloud
[Benioff:2009]
• SIEBEL enterprise sales force management $
1. Price $1,500 per seat, at 200 users = 300,000
2. $54,000 for support (18%) /year, x 5 = 270,000
3. $1,200,000 consulting for installation =1,200,000
4. $100,000 admin.personnel/year, x 6 = 600,000
5. $ 30,000 training / year, x 6 = 180,000
 6 years’ usage Total = 2,550,000
Note that the customer’s total is >> than the price
10/28/2011 CS207 16

17. Cloud delivery by
salesforce.com
• Benioff Saleforce.com new entry:
 $150.-month & user only -- monthly billing
 Make interface look like Amazon – no training needed
 Low risk for individual adopters
 Still a high risk for a changeover in large businesses, where
changes are controlled by a risk-adverse IT manager or CIO.
 Start focusing on small businesses
 Hard to reach a broad market with little cash
 Must make a lot of noise
 Later sales force had to change its initial model
 Deal with large companies
 Deal with the Dot-com bust, when many companies failed
Business must remain flexible
10/28/2011 CS207 17

18. Advertising
1. Audience
3. Logo & name
Focused
Essential for branding
 Salesforce
In front of competitors Metaphor
4. Timing
annual sale meetings 3x
1. Fake demonstrators in SF. Have Product ready
2. Give coffee, mugs, rides, • Few bugs
literature to attendees in NY • Clear operation
3. Hire all taxis in Nice, give Negative? • Useful
free rides to site in Cannes.
Vs. Superbowl? 2. Address
• Much buzz a. Buyers in corporations c. Both
• Huge audience b. Users and employees
• Your audience? • Understand motivations for change
10/28/2011 CS207 18

19. Review Allocation
• When is allocation needed?
1. Tech. , Pharma company:
 income due to R&D versus advertising
2. Financial Company:
 income due to software versus investment experts
3. Internal ▬ product mix
• For the Pareto-optimality allocation of income we
use cost.
 But recall: Do NOT use cost as a surrogate for value,
value of intangibles come from derived income.
28-Oct-11 CS207 19

20. Expense Rollover
A valuation based on cost
1. Collect the expenses ei over the total lag period p
2. Adjust the expenses by a discount rate d, ai= (1+d)p-i
3. For year i = 1 → p estimate the R&D retained ri =1- 1/p
4. Aggregate retained to the end date, R = Σ ri x ei x ai
5. From experience, publications obtain an expected
expense to income margin m; m can range from 1 to 20 ...
m≈2 in the first model we used
6. Expected value of IP V = m x R
But the estimation of m is verrrrrrrrrrry iffy
Technological advances are rarely stable
But could be used for a) advertising -- much untrustworthy data
b) stable maintenance component only
28-Oct-11 tCS207 20

21. Setting License fees
Say you want to delegate sales in Europe to some
company EUsales that can do it easier over there
• How do you set the fees or royalties?
1. You have computed a value of your SW of $1M
 But without discounting, it is actually $1.6M = Σ(due old, slide 5)
 You will also maintain the SW 1.36M = Σ(maintenance cost, slide 12)
The total due is $3M
2. You expect the European sales will be 40% of total, 20 000
 The reason for not discounting is that funds arrive at the same times.
• To earn the same you should charge 1./2.= $150/unit
 It does not matter how EUsales sells it and what it charges
 Complexities are required language, interface improvements
10/28/2011 Gio: CS207 21

22. Discussion
• A long-lived product is hard to displace if
 It is well maintained,
but that becomes costly
 Keeps up with all standards
• Internal replacement
 Should be easier
 But has not been in practice
Next week IP, IP protection, and IP licensing
28-Oct-11 CS207 22