January 8th, 2001

Decision Hierarchy Structure

MODEL: CDP Roll Out Prioritization

Notes:

This is the generic rollout priority model for Product X development.

Its purpose is to prioritize features and bugs for inclusion in a particular rollout.

Model Methodology: Simple Multi-Attribute Rating Technique

CDP Analysis - Hierarchy Model Profile

Hierarchy Background Information

In the decision hierarchy you systematically identify all relations that will require quantification: subcriteria whose relative importance is to be determined with respect to main criteria, alternatives to be measured against lowest criteria. As you consider what is required to discriminate between the various alternatives, you can define new relationships between criteria, add or delete criteria, alternatives and levels.

Contributions by Criteria Comments

It is clear that the ranking of the features/bug fixes is almost completely determined by high "market necessity" and low "costs".

Background Information

Contributions By Criteria - Stacked Bar Chart View Background

In the stacked bar chart view, you can see how the various criteria in any given level contribute to the alternatives' decision scores. It is a useful screen for checking that the decision model is reasonable in that alternatives that you expect to do well, given their high scores against the more heavily weighted criteria, do indeed to well.

Contributions By Criteria - Radar Graph View Comments

It is clear that the subcriteria "design priority", "develop hours", "bug priority" and "User Demand" dominate the contributions to the rankings of the features/bugs. This reflects the low importance assigned to the main criteria of "difficulty" and "desirability". While we may still argue over the relative importance of "desirability", the (relatively) low importance of "difficulty" is dangerous given the tight deadline we have to finish the release.

Background Information

Contributions - Radar Chart View Background

The Radar Chart view of the Contributions By Criteria chart makes it very clear which lowest criteria, by dint both of their relatively high weights, and high scores of the alternatives against them, truly have the largest input to the outcome of the model.

Tradeoffs Comments Here

Many of the lowest criteria have qualitative scales, which makes their tradeoff values hard to interpret. Development Hours are the key resource, and all other criteria are measured in terms of them. Obviously one could ascribe a dollar value per development hour, but since we have a finite number of developers, the key resource constraint is the total number of developer hours available.

CDP Analysis - Comments on Tradeoffs

The Reference criterion is "develop hours", and tradeoffs are specific to the highest scoring alternative, "Win 95 Modal".

In terms of the lowest criteria with numeric scales,

1 hour of develop is being traded off at 2.0 testing hours

1 hour of develop is being traded off at 1.28 repro hours

1 hour of develop is being traded off at -0.06 times requested

Background Information

In any difficult decision, many of the criteria conflict - you desire the highest quality at the lowest price. The essence of the multicriteria decision is the tradeoff between criteria. For the lowest criteria (or attributes), these tradeoffs can often be explicitly calculated in terms of the scale units of those criteria. Comparing these tradeoffs against internal or industry standards is key to validating the decision model.

As was agreed at the meeting on January 5th, 2001, we can only have two programmers (full time) to finish the release. If the code freeze deadline is still three weeks from now, that only gives 260 work hours available. As the database assumes pure development hours, and we agreed each such "develop hour" corresponds to two physical work hours, we only have 130 develop hours available.

From the cumulative total of "develop hours" shown, we will only get down to the "Print Zoom Disable" bug. The red bars above indicate that that feature has already been put off to the next release, as was hammered out over the last week, or the bug could not be reproduced. As usual, we will generally work from the top priority features/bugs such as "WinModal 95" down, so we may not get to "Blank Column Reports", etc.

Decision Scores Background - Portfolio

When the decision task is to select more than a single alternative, you may wish to choose as many of the highest scoring alternatives as your resources allow. For instance, if each alternative has a cost, and you have a budget, you may only be able to afford the first 5 items in your prioritized list.

The Decision Scores window in DecisionPlus is a powerful tool for determining which are the best alternatives your limited resource can afford. By sorting the results by their desirability, then showing the cumulative amount of your resource required to afford all alternatives, starting with the most desirable, you can clearly determine where your budget would be exhausted.

We can, with two full time developers, meet the specified code freeze date. However it is likely that we will not include all those features/bug fixes that were designated for this release. Now if we could have one more week....