Part 1 identified available tools for project portfolio management. Part 2 described key differences. Part 3 summarized costs and risks. Part 4 identified inability to identify optimal project portfolios as the weak link for most tools. Part 5 described decision models, and Part 6 provided criteria for evaluating tools. This final part summarizes recommendations for those who are purchasing or building project portfolio management tools.
Tools are good. Tools for project portfolio management can help organizations improve the selection and management of their project portfolios. This can and will allow organizations to increase value even while cutting costs. Tools promote a more deliberate, careful, and consistent evaluation of project alternatives. They force the generation of more and generally better project data. Many of the available tools provide excellent data management and reporting capabilities, making it much easier for managers and executives throughout the organization to understand the work that is being conducted.
All tools, however, are not equally good.
Which Approach is Best?
According to Thomas Davenport and Jeanne Harris, "Senior executives' top priority for their systems is obtaining improved information for decision making." The authors' studies show that the most successful companies make the greatest use of quantitative analysis. "High performers recognize the importance of using enterprise systems for better insights. They were five times as likely as low performers to report that analytically driven insights are a key element of their business strategy." 
If you've read the rest of this paper, you already know my answer. The best tool is one that actually helps your organization make better project choices. Outputs that support the management of individual projects are nice, but the greatest opportunity is the ability to use analysis to enable the organization to make the best project choices. For a tool to do this, it must make the right recommendations, and this requires that it be based on a decision model that is accurate, logically sound, and complete. Furthermore, the tool must be practical, effective, and acceptable, otherwise it won't be used or its use won't actually influence decisions. Although some tools are clearly better than others, no one approach is best (or even adequate) for all circumstances. The choice of a PPM tool needs to be made differently for different organizations and applications.
Get Quality Analytics
"Often companies err by focusing on getting software installed, but they miss the opportunity to get the analytics and the forward looking information enterprise systems can provide." (Kevin Carnahan, managing director of system integration for Accenture) 
More specifically, for a tool to be the best approach, it must compute project priorities based on the value that alternative project portfolios would provide. An acceptable tool must be based on sound project valuation theory, one that captures best understanding of how projects actually create value for the enterprise. As described in the previous parts of this paper, most tools aren't based on sound project valuation theory and, therefore, fail to provide accurate estimates of project portfolio value. As a result, they do not offer much help for making tough project choices. The quality of the model used to quantify project and portfolio value, in my opinion, is the critical discriminator for choosing a PPM tool.
More than 100 tools are currently being marketed for project portfolio management (see the list in Part 1).
Custom-Designed vs. Configurable Tools
A significant difference among tool providers is one of philosophy. As noted previously, the approach favored by most software vendors and many consultants involves creating a tool based on a "configurable" decision model — a model that is hard-coded within the software but includes parameters and options that can be set to help fit the tool to a range of different applications or situations. Frequently, marketing materials describe such tools as being "fully customizable," but the truth is they can be adjusted only within a narrow range allowed by model parameters.
The alternative approach, favored by some consultants, is one that provides flexibility for creating custom decision models. These tools are created on modeling platforms — high-level programming languages designed to facilitate general purpose modeling and analysis. These platforms include Analytica, Excel, Visual Basic, Crystal BallŪ, and DPL Portfolio, as well as web-based, project portfolio management tools that allow one of these or a similar modeling platform to be accessed via a web portal.
Each approach has its own advantages and disadvantages. Tools with configurable models are convenient. They can be implemented quickly. They do not require effort on your part to design a custom model—that work has already been done. They key question is whether the configurable model is adequate for the application. Potentially relevant questions include: Does the tool provide capability to handle all types of projects and project portfolios that might eventually need to be analyzed? Does it account for all types of project benefits, including dynamic, time-varying project impacts? Does it allow for rigorous multi-attribute utility valuation of projects? Does it allow for true portfolio optimization (not just ranking)? Does it allow for risk valuation as well as risk characterization?
Configurable models are typically programmed in software languages that are not friendly to changes. If the configurable model doesn't provide some capability (either a capability currently desired or one that might be desired in the future), beware that it may be difficult or impossible to obtain this capability. The code may be so complicated that only the original programmers are capable of making changes. Changes that affect structure often produce ripple effects that require extensive rewriting of source code. For example, if a tool expects cost savings resulting from a project to be entered as an annual average value, changing the model to allow entering year-by-year estimates can be difficult.
Custom tools built on modeling platforms are much more flexible than tools based on pre-set, but configurable models. (Compare the ease of changing an Excel spreadsheet with the difficulty of getting a software vendor to make a model change to its tool.) Custom tools implemented on modeling platforms can more easily "grow" as the user organization gains experience and understanding. However, developing customized, quality tools for project portfolio management using Excel or another general-purpose modeling platform can be labor intensive. It requires the client organization to participate in the design process by making choices and requires a consultant skilled in modeling and portfolio analysis to implement the custom model on the modeling platform. The process is faster if the consultant has compiled a library of sub-models for use as building blocks, and if the tool automates some of the labor-intensive programming steps (such as creating user input templates for the custom model). However, tools with flexible modeling platforms can be large and more costly to develop, so they may be more expensive to acquire than configurable tools.
If a configurable tool fits the need and contains a defensible logic for valuing projects and optimizing project portfolios, then such a tool can probably be implemented more quickly and with less cost than a custom tool. However, these are big "if's." My experience to date has been that customers who want a tool that correctly identifies value-maximizing project portfolios (in situations where project value is more than discounted project cash flow) must either use a custom tool built on a general modeling platform or must link a vendor tool (i.e., tool providing the desired data management and reporting capability) to external models that correctly compute non-financial components of project value. However, tool capabilities are advancing rapidly, and new options will become available that reduce the costs and difficulty of obtaining a tool that meets all of the needs of the organization.