Reference no: EM131135668

Decision Analysis Model - PeachyKeen Software

THE SCENARIO It is 5 years from now, and you work for PeachyKeen Software as Assistant Vise President for Long Range Planning. PeachyKeen is not truly a software house, but rather a software publisher, and its business is based on the following sequence of events:

• An independent developer writes an innovative software package. The developer approaches PeachyKeen to distribute the software. PeachyKeen considers the intended function of the software and what it knows of the developer's skills. If they see little or no hope, they reject the project. Otherwise, the move to the next step. PeachyKeen tests the new software for bugginess, user interface, usefulness, features, and other areas of concern. This testing costs are just a routine cost of doing business. If the new software passes testing, PeachyKeen and the developer negotiate. If they strike a deal, then PeachyKeen specifies final "polish" (including documentation), and arranges for distribution disks, manuals, packaging, and distribution.

Your boss, Kay Smith, the V.P. of Long Range Planning, has just met with John Werkin, who has developed a new speech-processing package named Ear. With a cheap telephone handset or headset (not provided), Ear will perform most functions of a keyboard and mouse. Ear has passed all of PeachyKeen's tests with few glitches. Kay is excited about the prospects for Ear, and wants you to help her decide whether the company should publish this product.

Kay and John have worked out a tentative deal. If PeachyKeen publishes Ear, it will pay John $300,000 plus 5% of gross sales. While a product like Ear will carry much higher retail price, PeachyKeen would actually sell it into its distribution channels at $25 per copy for large orders, with the buyer paying all freight costs. PeachyKeen would calculate John's 5% based on the $25 per unit of actual revenue.

Kay's best estimate is that, after John does a final cleanup of his software and manuals, it will cost PeachyKeen about $150,000 to edit the manuals, set type, create master installation disks, design packaging, advertise, and generally prepare the product for the manufacturing stage of publishing. If they cancel the deal at now, of course, they avoid these costs.

PeachyKeen's manufacturing process is set up in such a way that there is always a manually produced test run of 100 copies first. After the test run, the process is set up for a run of any number of copies. All later batches of the software must be made in that number. It costs $25,000 to set up a batch run (even the test run). For Ear, the variable manufacturing cost is $10 per copy manufactured.

Although it is a great oversimplification, Kay is willing to make the decision based on the possibility that Ear will be either a Dog, a Standard, or a Killer. A Dog usually loses money. A company usually hopes that most of its new products will prove to be Standards. A Killer application, if handled right, is cause for great celebration.

2

In this case, she defines a Dog as selling only 10,000 copies in the first year. A Standard would sell 50,000 copies. If it is a Killer application, Ear will sell 250,000 copies in the first year on the market. It is PeachyKeen Software's policy that they will meet the demand. If they plan, for example, on a product being a Standard and it turns out to be a Killer, they will produce as many more batches as may be required to meet demand, even if the extra setup costs are painful. Since the possible outcomes appear to be a Dog, Standard, or Killer application, Kay thinks it only makes sense to consider batch sizes that correspond.

PeachyKeen does not have to waste the copies made in the test run. They could send them to a variety of publications as review copies at the total cost of $10,000. The sequence of the decisions Kay needs to make and the timing of these events is as follows:

• First Kay makes a decision to take John's deal. Although this is a decision she should be contemplating as well as any other, we will assume, for simplicity, that this decision will be made.
• Then she needs to spend time and money to get to the point of performing test runs. After the test run copies have been produced, she has to decide whether to send them out for reviews. If she decides to send them for reviews she will consider the results, and then make one of the following decisions: (1) drop the product from consideration, (2) make Ear in batches of 10,000, (3) make Ear in batches of 50,000, or (4) make Ear in batches of 250,000.

Kay, after considering historical review data for the similar software products, has come up with the following reliabilities and prior probabilities:

Reliabilities Dog Standard Killer Rave 0 0.12 0.67 Good 0.43 0.84 0.33 Poor 0.57 0.04 0 Priors 0.14 0.50 0.36

If Kay decides not to send the copies for reviews and goes directly to production, she needs to decide on the batch size only. In considering the Ear decision, keep in mind that making 10,000, 50,000, or 250,000 copies per batch is a decision to which PeachyKeen must commit before they know whether the product will turn out to be a Dog, Standard, or Killer application, but after reading the reviews (if they decide to have the test copies reviewed). Not only that, but the reviews do not determine the demand levels. They only improve your state of information about what demand levels might materialize.

Kay has some very specific questions about the possible sequence of decisions, as well as the expectations on the project return:

1. Assuming the reviews were conducted for $10,000, if the results are Poor, should PeachyKeen abandon the project? Produce in batch sizes of 10,000? 50,000? 250,000? What about if the reviews are Good? Rave?

2. In each of those cases, what is the best course of action based on maximizing Expected Return (ER). What is the ER for the project as a whole?

3. What is the Expected Value of Sample Information (EVSI)? Should Kay have the reviews conducted for the price of $30,000? $50,000? What is the most Kay should be willing to pay for the reviews?

YOUR JOB This is a task that calls for Decision Analysis model in a spreadsheet format. It is critical that you follow the principles of effective spreadsheet development that we have stressed in class. Kay could change her mind about how many copies of a Dog, a Standard, or a Killer the company might sell. Or she might get new cost estimates, or negotiate a better deal with John Werkin. She'll want to be able to change all those kinds of things easily and still get meaningful results.

The easiest way to start is with a blank Excel workbook, named using our standard convention (p2s5044*g**.xls). Continue by designing and implementing the profit model in a single worksheet named pk_software. The top-left corner cell of pk_software worksheet must contain the heading: Decision Analysis Model - PeachyKeen Software, in bold and 14 point font. In the same worksheet, you must have the Payoff Table for each combination of a decision alternative (e.g., batch size of 50,000) and a state of nature (e.g., product is a Killer application), in a single table calculated from the profit line using Excel's Data Table function.

Your main task is to implement the solution to PeachyKeen's decision problem with the decision tree. Before creating the tree, however, you need to calculate all the required probabilities (marginal and posterior) using the given reliabilities and priors. Enter the table with reliabilities and prior probabilities from the previous page below the payoff table. Below this table set up another table for marginal and posterior probabilities. These probabilities must be calculated using the formulas discussed in class, and formatted to two digits. In other words, absolute and relative referencing must be used for all the probabilities, as well as SUMPRODUCT function for the marginals.

The decision tree must be created in a separate worksheet, named dec_tree using Exel's add-in TreePlan. You must correctly specify the decision and event nodes, branches with description of the decisions (Review, No Review, Drop It, Batch 10K, Batch 50K, Batch 250K), outcomes of the reviews (Rave, Good, Poor), and the states of nature (Dog, Standard, Killer). All the numeric values for the decision tree in dec_tree sheet must refer back to appropriate tables in pk_software worksheet.

All the answers to Kay's questions must be readable from the decision tree. All of the spreadsheet techniques you have learned this semester must be incorporated. That includes (but is not limited to) no hard-coded parameters, clearly labeled variables, accurate results, smooth flow of the model from a user perspective, and appropriate formatting (numbers, borders, colors, titles at the top of the sheet, sheet names, etc.).

In addition to the model, give Kay a brief memo stating your recommendations by incorporating the answers to the above three questions. In addition to batch size recommendation and justification in terms of ER, you must instruct Kay on how to interpret the decision tree, and provide guidance on how to use of the entire model. Embed the memo in the sheet with the profit model and payoff table (pk_software). Use To: Kay Smith, From: Group Member Names, and Subject: PeechyKeen Decision Analysis as your header.

The last worksheet in your model, named grading, must be the grading checklist (without the points) that lists all the elements you will be graded on.

Model Accurate profit model 2 Accurate payoff table 1 Accurate probabilities 1 No hard-coded parameters 1 Good user interface 1 Decision Correct layout 1 Tree Accurate values and results 1 Memo Questions answered 2 Total 10