barry has entered the room . Techsupport has entered the room - barry: Good morning, everybody. We got a lot of good questions. I'll start entering and answering them now. 1 barry: Q: What should you do if your constraint set leaves you with an empty set of feasible solutions? 2 barry: 1 Reconvene your stakeholders, summarize the situation, and see where they may be willing to relax the constraints. If the stakeholders can't agree on relaxing any of the constraints, indicate that you're going to cancel the project. This sometimes will get stakeholders to relax constraints a bit. . temp1 has entered the room 3 barry: Q: Given the tremendous impact of people on software productivity, how sensitive are models to changes in people factors? Any books good to read for further guidance on software people factors? 4 barry: 3 Not only COCOMO(TM) II, but also, COCOTS, CORADMO, COQUALMO, and most other commercial software cost estimation models have people factors as their highest-leverage variables (except for the size of the product, which also translates into relative size of the project team). The best recent reference is Version 2.0 of the People Capability Maturity Model (July 2001), available at http://www.sei.cmu.edu/cmm-p/ It has numerous references to other books, reports, and papers. 5 barry: Q: Since all the decision rules under complete uncertainty (maximin, maximax, Laplace), have pitfalls, does it make sense to try all three rules and understand their recommendations? 6 barry: 5 Yes. Often, you can determine sensitivities and subjective preferences by confronting a recommendation that just looks good or looks bad to you. 7 barry: Q: Can we determine the maximum number of constraints possible to get an optimal solution? 8 barry: 7 No. For example, in TPS II Option A, the minimum throughput constraint E>2350 will constrain the solution to N = 5 (or 6) processors and E = 2400. The constraints E>2000 and Av>.98 will constrain the solution to N = 4. The 50 constraints E>2350, E>2351, ..., E>2399 will still just select N = 5 (or 6). 9 barry: Q: What is an example of a curvilinear constraint? 10 barry: 9 A good example is the COCOMO(TM) II mionimum-schedule constraint, which says that for a given project effort in person-months, feasible project schedules in calendar months are constrained to values higher than about 2.5*(PM)**(1/3) 11 barry: Q: What are someexamples of non-managerial utility functions? 12 barry: 11 One is gambling at Las Vegas. Everyone knows that your expected outcome is negative, but for some people, the rush they get from putting their money at risk and the occasional winning day is worth it. Other exaples might be bungee-jumping and rockclimbing. 13 barry: Q: What are the side benefits from building prototypes. Are there counterpart drawbacks? 14 barry: 13 Some side benefits are the marketing benefits of having somethying to show to potential customers, and the opportunity to get greater stakeholder buy-in to your project by having them experience the potential benefits of the product, and having them participate in its definition and refinement. A counterpart drawback is the risk that impatient customers or users will take the flaky prototype and put it into service, leaving you with a lot of usage complaints and rework. 15 barry: Omer -- I just clicked on Threaded Agenda and got last week's agenda, even after hitting Refresh. Please fix this. 21 Techsupport: 15-professor i'll have this fixed later in the day as i do not have access to them in real time 22 Techsupport: 15-you can still get this week's log by scrolling down though 16 barry: Q: Which risk behavior is best: risk-averse, risk- indifferent, or risk-seeking? 17 barry: 16 The best behavior is situation-dependent. Risk-indifferent is generally not good, but in a low-risk situation it can be better than delays due to risk-averseness or making the situation riskier due to risk-seeking behavior. In a startup company, risk-seeking is generally better (up to a point); once the startup company has a large established base of customers and core services, a risk-averse approach to the core services is best, although risk-seeking behavior in generating new capabilities is still best. 18 barry: Q: Is loss-avoidance a good goal to have your employees strive for? 19 barry: 18 It's best to have categories of situations in which loss-avoidance is encouraged or discouraged. Creating losses via product gold-plating or bureaucratic overhead-padding should be discouraged. But for new-product research and development, loss-avoidance should be discouraged. When I ran the Information Science and Tecvhnology Office at DARPA, the Defense Department's equivalent of a venture capitalist organization, having a 100% project success rate in your office was considered too conservative and bad. 20 barry: Q: Does an optimal solution in a business case always yield the maximum profit? 23 barry: 20 Not always. For a fixed-period business case, say 5 years, the maximum-profit solution might turn the organization into a "cash cow," and neglect spending on new features to keep profits high after 5 years. [Ref: 21,22]24 barry: 21,22 OK; thanks! 25 barry: Q: How can we use the Expected Value of Perfect Information (EVPI) to help make decisions? 26 barry: 25 One example is that if the EVPI is very small, this means that there is not much to be gained by further analysis, and that any of the alternatives will be about equally good to choose. 27 barry: Q: Do Extreme Programming and Linear Programming relate to each other? 28 barry: 27 Not really. XP is a very qualitative approach to building software. LP is a very quantitative approach to making decisions. In the 1940's, "programming" related to making resource allocation and scheduling decisions, which is how linear programming, dynamic programming, etc., got their names. 29 barry: Well, this has been a very disappointing session, with no Q&A dialogue at all. I will be sending out a query to everybody, but particularly to the DEN students, asking if the chat room sessions are worth contiuing and if so, how to get more participation going on in them? Everybody please consider this.