Week 7

advertisement

INFO 631

Prof. Glenn Booker

Week 7 – Chapters 19-20

INFO631 Week 7 1 www.ischool.drexel.edu

Break-Even Analysis

Ch. 19

Slides adapted from Steve Tockey – Return on Software

INFO631 Week 7 2 www.ischool.drexel.edu

Present Economy

• “Present Economy”: decision techniques not involving time-value of money

– Decision made based on 1 or more decision variables and 1 or more objective functions

– Techniques:

• Break-even (Ch 19)

– Finds value of the decision variable where performance is identical between alternatives

• Optimization (Ch 20)

– Finds value of the decision variable(s) with the best performance.

INFO631 Week 7 3 www.ischool.drexel.edu

Break-Even Analysis

Outline

• The situation

• Decision variables and objective functions

• Break-even with two alternatives

• Break-even with three alternatives

• General case break-even analysis

INFO631 Week 7 4 www.ischool.drexel.edu

The Situation – Chicago

Company

• Chicago-based software project team needs .Net training but hasn’t decided how many people need it

• Team finds reputable Los Angeles-based training company

• Chicago project manager has two options

– Send people to LA for training

• Cost is $1620 per person for tuition, travel, expenses, …

– Hire instructor to come to Chicago

• Cost is $17,975 including fee, instructor travel, & expenses

• At what point is it better to have the instructor come to Chicago instead of sending team members to LA?

• Essence of break-even analysis: knowing the break-even point, deciding is real easy

• Question to answer: You’re the manager of a Chicago-based SW team that needs this training. How to best use your project $?

INFO631 Week 7 5 www.ischool.drexel.edu

Decision Variables and Objective Functions

• Decision variable

– Set of possible values for some choice in a decision analysis

• E.g., the number of people that get .Net training

INFO631 Week 7 6 www.ischool.drexel.edu

Decision Variables and Objective Functions

• Objective function

– Equation relating values of the decision variable to performance of an alternative

• CostInLA = $1620 * #People

• CostInChicago = $17,975

– Types

• Income function:

– Relates values of decision variables to income

– Income function  income

• Cost Function

– Relates values of decision variables to cost

– Cost function  cost

INFO631 Week 7 7 www.ischool.drexel.edu

Break-Even With Two Alternatives

• Find value of decision variable where objective functions are equal

• Algebraic solution

– Set the objective functions equal to each other and solve for the decision variable

INFO631 Week 7 8 www.ischool.drexel.edu

Break-Even With Two Alternatives

– Example

• Know CostInLA = $1620 * #People and CostInChicago = $17,975

• Set CostInLA = CostInChicago

• $1620 * #People = $17,975

• #People = $17,975 / $1620 = 11.1

• So what does this mean:

• If <= 11 people need training, send them to LA

• If > 11 people need training, have instructor come to Chicago

INFO631 Week 7 9 www.ischool.drexel.edu

Break-Even With Two Alternatives

• Graphical solution

– Plot the objective functions and find the intersection

LA

$20K

Chicago

Cost

$10K Break-even point

$0

#People

INFO631 Week 7 10 www.ischool.drexel.edu

Break-Even With Three Alternatives

• Adding another alternative makes the analysis more complicated

• But basically still straight forward

– Graphical

– Algebraic Solution

INFO631 Week 7 11 www.ischool.drexel.edu

Example - Add option to

Chicago Company

• New Option for Chicago Company:

– LA training company makes an offer involving a session in Denver

• CostInDenver = $5000 + $925 * #People

– To Solve

• Easiest to see graphically

– See next slide

– Always look for the solution with the lowest cost!

INFO631 Week 7 12 www.ischool.drexel.edu

Break-Even With Three Alternatives

• Graphical Solution – New Option Added

CostInDenver = $5000 + $925 * #People

Break-even

Denver-LA

LA

$20K

Chicago

Cost

$10K

Denver

Break-even

Denver-

Chicago

$0

#People

INFO631 Week 7 13 www.ischool.drexel.edu

Break-Even With Three

Alternatives

• How interpret?

• New option is viable for a middle case

– If <= 7 people need training, send them to LA

– If 8 to 13 people need training, send them to

Denver

– If > 14 people need training, have instructor come to Chicago

INFO631 Week 7 14 www.ischool.drexel.edu

Break-Even With Three Alternatives

• If the Denver option were

– Unlikely, but possible--same break-even as Chicago:LA,

Denver only viable at exactly 11 people

CostInDenver = $10,000 + $725 * #People

LA

$20K

Chicago

Cost

$10K Denver

$0

#People

INFO631 Week 7 15 www.ischool.drexel.edu

Break-Even With Three Alternatives

• If the Denver option were

– CostInDenver = $7,500 + $1300 * #People

– Also a possibility: Denver always worse than LA or

Chicago, it’s never viable

$20K

Chicago

Cost

$10K

Denver

LA

$0

#People

INFO631 Week 7 16 www.ischool.drexel.edu

Break-Even With Three Alternatives,

Algebraic Solution

• Solve LA:Chicago

– CostInLA = $1620 * #People CostInChicago = $17,975

– CostInLA = CostInChicago

– $1620 * #People = $17,975

– #People = $17,975 / $1620 = 11

• Is Denver better or worse than

LA:Chicago?

– Cost at LA:Chicago break-even (11 people) is $17,975

– CostInDenver for 11 people = $5000 + $925 * 11 =

$15,175

– Denver is better than LA:Chicago, discard LA:Chicago

INFO631 Week 7 17 www.ischool.drexel.edu

Break-Even With Three Alternatives,

Algebraic Solution

• Solve LA:Denver

CostInLA = $1620 * #People CostInDenver = $5,000 + $925* #People

CostInLA = CostInDenver

$1620 * #People = $5,000 + $925 * #People

($1620 * #People)

– ($925 * #People) = $5,000

$695 * #People = $5000

#People = $5000 / $695 = 7

• Is Chicago better or worse than

LA:Denver?

Cost at LA:Denver break-even (7 people) is $11,475

CostInChicago for 11 people = $17,975

LA:Denver is better than Chicago, keep LA:Denver

INFO631 Week 7 18 www.ischool.drexel.edu

Break-Even With Three Alternatives,

Algebraic Solution

• Solve Chicago:Denver

CostInChicago = $17,975 CostInDenver = $5,000 + $925* #People

CostInChicago = CostInDenver

$17,975 = $5,000 + $925 * #People

$12,975 = $925 * #People

#People = $12,975 / $925 = 14

• Is LA better or worse than

Chicago:Denver?

Cost at Chicago:Denver break-even (14 people) is $17,975

CostInLA for 14 people = $22,680

Chicago:Denver is better than LA, keep Chicago:Denver

INFO631 Week 7 19 www.ischool.drexel.edu

Break-Even With Three Alternatives,

Algebraic Solution

• Sort the valid break-even points by increasing decision variable value

At 7 people, LA:Denver break-even

At 14 people, Chicago:Denver break-even

• Reason about the segments

Denver must be best between 7 and 14 people, so LA must be best below 7, and Chicago must be best above 14

INFO631 Week 7 20 www.ischool.drexel.edu

General Case Break-Even

1. Calculate the break-even point for each pair of objective functions

1. With n functions, there will be (n*(n-1))/2 candidates

2. Discard all break-even points that are:

1. Dominated by any other objective function at that value of the decision variable

2. Outside the reasonable range of the decision variable

(e.g., too-low values and too-high values)

3. Sort the remaining, valid break-even points in order of increasing decision variable value

4. Reason about the segments

1. When the same objective function is in consecutive break-even points, it’s the best between those points

INFO631 Week 7 21 www.ischool.drexel.edu

Key Points

• Break-even analysis chooses between two or more alternatives by figuring out which points, if any, would be indifferent between those alternatives

• A decision variable represents a set of possible values for some choice

• An objective function is an equation relating values of decision variables to performance of an alternative

• To find break-even points with two alternatives, set the objectives functions equal to each other and solve for the value of the decision variable where that happens

– The graphical approach finds the intersection on a graph of the objective functions

• With three or more alternatives, break-even points between each pair need to be considered

– Some points may be dominated by other alternatives and will need to be discarded

INFO631 Week 7 22 www.ischool.drexel.edu

Optimization Analysis

Ch. 20

Slides adapted from Steve Tockey – Return on Software

INFO631 Week 7 23 www.ischool.drexel.edu

Optimization Analysis

Outline

• Introducing optimization

• One alternative, one decision variable

• Multiple alternatives, one decision variable

• One alternative, multiple decision variables

• Multiple alternatives, multiple decision variables

INFO631 Week 7 24 www.ischool.drexel.edu

Introducing Optimization

See also Ch 11 – Economic Life - Graph

• Find the point where overall performance is most favorable

• Useful when an objective function has 2 or more competing components

– One component increases with the decision variable

– Other component decreases

Economic life

AE(i)

Costs

Total cost

Operating and maintenance cost

Capital recovery with return, CR(i)

Number of years the asset is kept

INFO631 Week 7 25 www.ischool.drexel.edu

Introducing Optimization

• Can be applied to maximizing an income function

– Finding max point on an income function rather then min point on a cost function

– Use same techniques

• Just look for min point rather than max point

INFO631 Week 7 26 www.ischool.drexel.edu

Introducing Optimization

• Two ways to solve

– Algebraic (elegant) and

• Uses differential calculus

– Min/Max when 1 st derivative = 0

– Graphical (brute-force)

• Run multiple sample values for decision variable through function and narrow in on best result

• Methods

– One Alternative, One Decision Variable

• Simplest

– Multiple Alternatives, Single Decision Variable

– Single Alternative, Multiple Decision Variables

– Multiple Alternatives, Multiple Decision Variables

INFO631 Week 7 27 www.ischool.drexel.edu

One Alternative, One Decision Variable

• Performance = F ( Decision variable )

• Two ways to solve

– Algebraic (elegant) and

• Uses differential calculus

– Min/Max when 1 st derivative = 0

– Graphical (brute-force)

• Run multiple sample values for decision variable through function and narrow in on best result

INFO631 Week 7 28 www.ischool.drexel.edu

One Alternative, One Decision Variable

Example: distributed application needs queuing buffer

– More messages/packet reduces network overhead but increases average queue time per message

– Overall queuing delay described by the cost function:

TD

18r

648

21 r

• TD is total delay in milliseconds

• r is number of messages queued per packet

• 18r component is average queuing delay

• 648/r component is variable network overhead

• 21 is fixed network overhead

– What is optimum messages/packet?

INFO631 Week 7 29 www.ischool.drexel.edu

One Alternative, One Decision Variable:

Algebraic Solution

• Find the first derivative dTD dr

18

648 r 2

• Set the first derivative equal to zero and solve for the decision variable

(on next slide) dTD dr

18

648 r

2

0

• The optimum messages/packet = 6

INFO631 Week 7 30 www.ischool.drexel.edu

One Alternative, One Decision

Variable: Algebraic Solution

• solve for the decision variable r

648

6

18

• The optimum messages/packet = 6 and

TD = 237 ms

TD

18 * 6

648

21

237

6

INFO631 Week 7 31 www.ischool.drexel.edu

One Alternative, One Decision Variable:

Graphical Solution

• Graphical solution (brute force)

– Plot the objective function, then find maximum/minimum point

– Note: In general, there can be several local min/max points on function o r m a

P e r f n c e

Overall minimum

P

Overall maximum

Q R

Decision variable

S

Overall:

Min = P

Max = Q

R & S = local min/max

INFO631 Week 7 32 www.ischool.drexel.edu

One Alternative, One Decision Variable:

Graphical Solution r TD

1 687

2 381

3 291

4 255

5 241

6 237

7 240

8 246

9 255

10 266

11 278

12 291

13 305

14 319

15 334

700

600

500

400

300

200

100

0 r

R = 6

TD

INFO631 Week 7 33 www.ischool.drexel.edu

Multiple Alternatives, Single Decision

Variable

• Multiple Alternatives and their functions are driven by the same, single decision

– Cost

1

– Cost

2

– …..

= F

1

( Decision variable )

= F

2

( Decision variable )

INFO631 Week 7 34 www.ischool.drexel.edu

Multiple Alternatives, Single Decision

Variable

• Find where each alternative has optimum performance

– Use single alternative, single decision variable techniques

• Use Graphical (brute force) or Algebraic (elegant)

• General Approach

– Find the performance of each alternative at its optimum point

– Select the alternative with the best performance at its optimum point

– Choose alternate with best value at optimal point

INFO631 Week 7 35 www.ischool.drexel.edu

Multiple Alternatives, Single Decision

Variable

• Graphical solution

Optimum point for A1

Alternative A1 has optimum point at P

Alternative A2 has optimum point at Q

A2 at Q is cheaper than

A1 at P, so choose A2 and run it at Q

$

Cost for A1

Optimum point for A2

Cost for A2

P Q

Decision variable

INFO631 Week 7 36 www.ischool.drexel.edu

Single Alternative, Multiple Decision

Variables

• Performance = F ( v1, v2, … )

– V1 = variable 1

– V2 = variable 2, Etc.

• Algebraic (elegant) solution

– Use multiple differentiation

• Differential calculus

• Graphical (brute force) solution

– Plot the surface and look for extremes

– Systematic search algorithm

– Monte Carlo analysis (Ch 24)

INFO631 Week 7 37 www.ischool.drexel.edu

Systematic Search of Decision Variable

Space

LowestDV1 := some selected value of DecisionVariable1

LowestDV2 := some selected value of DecisionVariable2

LowestDV3 := some selected value of DecisionVariable3

LowestCost = CostFunction ( LowestDV1, LowestDV2, LowestDV3 ) while DV1 runs over the range of DecisionVariable1 while DV2 runs over the range of DecisionVariable2 while DV3 runs over the range of DecisionVariable3 if CostFunction ( DV1, DV2, DV3 ) < LowestCost then LowestCost = CostFunction ( DV1, DV2, DV3 )

LowestDV1 := DV1

LowestDV2 := DV2

LowestDV3 := DV3

INFO631 Week 7 38 www.ischool.drexel.edu

Systematic Search of Decision Variable

Space

• When systematic search completed

– Optimum point will be (close to)

• LowestDV1, LowestDV2, LowestDV3 and

• Have the LowestCost

• Not much help when search space is very big

– Too many combinations

– Use Monte Carlo analysis (Ch 24)

INFO631 Week 7 39 www.ischool.drexel.edu

Multiple Alternatives, Multiple Decision

Variables

• Multiple Alternatives and their functions are driven by the multiple decision variables

– Performance

1

– Performance

– ….

2

= F

1

= F

2

( v1, v2, … )

( v1, v2, … )

• V1 = variable 1

• V2 = variable 2, Etc.

INFO631 Week 7 40 www.ischool.drexel.edu

Multiple Alternatives, Multiple Decision

Variables

• Same approach as multiple alternatives, single decision variable

– Find where each alternative has optimum performance

• Use single alternative, multiple decision variable techniques

– Find the performance of each alternative at its optimum point

– Select the alternative with the best performance at its optimum point

INFO631 Week 7 41 www.ischool.drexel.edu

Key Points

• Optimization analysis is useful when objective functions have competing components: balance components to find where overall performance is best

• Two ways of finding optimum point on a single alternative with a single decision variable

– Algebraic uses differential calculus

– Graphical uses computed values

• To optimize more than one performance function with a single decision variable, first find optimum for each then select the one with best performance at its optimum point

• Two ways of finding optimum point on a single alternative with a multiple decision variables

– Algebraic uses differential calculus

– Graphical uses computed values

• Optimizing multiple performance functions with multiple decision variables is just like optimizing multiple alternatives with a single decision variable

INFO631 Week 7 42 www.ischool.drexel.edu

Download