OR WAIT null SECS
A major unresolved issue for the pharmaceutical industry in the 21st century is that few, if any, optimization techniques have found their way into the research portfolio arena.
A major unresolved issue for the pharmaceutical industry in the 21st century is that few, if any, optimization techniques have found their way into the research portfolio arena. That is surprising, for several reasons:
The industry is consolidating; there are fewer and fewer companies. Those that remain are increasing their R&D spending at double-digit annual percentage growth. This pattern suggests that pharma is becoming mature and, like all mature industries, more commodity oriented. As "commoditization" occurs, there is a need for increased quality through the use of systematic planning and benchmarking metrics.
This article discusses some of the current issues in the use of optimization techniques to manage an R&D portfolio, particularly some recent trends that make optimization techniques a more practical choice for companies.
Early in the development of total quality management concepts the quality-improvement pioneer W. Edwards Deming offered:
"Competent men [sic] in every position, from top management to the humblest worker, if they are doing their best, know all there is to know about their work except how to improve it. Help toward improvement can come only from some other kind of knowledge. Help may come from outside the company, or from better use of knowledge and skills already within the company, or both."
In the arena of the R&D portfolio, the pharmaceutical industry actually has a poor performance record. The need for improvement in portfolio management exists because "scientists traditionally have been reluctant to try new R&D management approaches." (See "In the Eye of the Storm," Pharm Exec, April 2003.)
The need for improved portfolio management was expressed by Pfizer’s research chief, Dr. Peter B. Corr, who "is trying to reduce the number of projects that fail later in development, when costs are escalating. And Pfizer is forging earlier and stronger links between researchers and marketing executives" (New York Times, September 8, 2002).
The essential elements of effective portfolio management are well documented. They include working as a team with marketing, sales, and finance to:
1. Quantitatively determine acceptable ranges for the key input factors of timing, cost, risk, and expected return (e.g., NPV) for each project under consideration.
2. Select projects based on management specified criteria aimed at balancing risk and return.
Tiggemann et al (in the Drug Information Journal, vol. 32, 1998) state, "the key is to focus on the best projects in terms of NPV, taking advantage of experience curve effects and then balance the portfolio accordingly to insure both short- and long-term success."
This is a bottom-up approach in the sense that, first, various graphical displays and spreadsheets are used to assess the value of the current portfolio. Then comparative results are determined for "trial" alternative portfolios to identify and exploit opportunities for improvement. This type of fine-tuning will lead to improvements in the strategic portfolio, but could provide results that fall short of a portfolio that is "optimal" with respect to minimizing risk and/or maximizing ROI. In addition, seeking improvements by adjusting the current portfolio may overlook opportunities that might be of significant long-term value to the firm.
Recent advances in portfolio management go beyond the current commonly used practice of identifying which projects to continue or drop based on consideration of a set of reasonable scenarios. Optimization modeling as applied to portfolio management is not new. For example, mutual funds use optimization techniques (e.g., "Markowitz" models) to select stocks that will maximize overall return subject to a desired overall risk as measured by volatility. However, approaches applied successfully to other disciplines are not directly applicable to pharmaceutical R&D portfolio optimization for a variety of reasons, including differences in risk measurement and the impact of multiple clinical phases.
The two charts pictured here illustrate two approaches to the problem of optimizing an R&D portfolio. The chart on top ("Structuring the Pipeline"), drawn from the research of Min Ding of Penn State University and John Eliashberg of the Wharton School, provides a good example of using modeling techniques to optimally structure the pipeline for a given therapeutic class. Ding and Eliashberg determine the number of projects at each clinical phase that maximize expected NPV, given information on the cost of development, probability of surviving each clinical phase and the magnitude of the business opportunity. Using historically based assumptions on required model input data, they show that achieving high NPV would require much greater levels of spending on more projects than in the past.
Two Approaches to Optimization
While these results are interesting and significant, they probably raise more questions than they answer. Can the number of projects actually be increased to levels anywhere near those suggested, given budget constraints and the current state of research? And if the number of projects can’t be raised to the levels suggested by the models, what is the best practical strategy? In particular, what number of projects should be in each therapeutic class to maximize portfolio NPV? These questions, and others, have been the subject of continued research and implementation of software to provide effective decision support to strategic R&D portfolio management.
The bottom chart ("Optimizing the R&D Portfolio") shows one response to these questions. Here the focus is not on maximizing NPV, but rather on getting the best possible NPV within a fixed budget. Where a pure NPV approach yielded results that couldn’t really be acted upon, these results (based on the actual research portfolio of a major pharma company) are fairly practical: Drop projects for immune systems, blood, and cardiovascular, and increase projects for cancer and infectious diseases. This result is the "top" of the top-down approach. Good decision-support software would allow R&D project management to evaluate changes in NPV for various "what if" scenarios relative to the initial optimal model results. These scenarios may be based on considerations of the firm’s own research, partnering, joint ventures, support of contract sales organizations (CSOs) and other avenues to augment their own acquisition/divestiture needs.
Is the pharmaceutical industry prepared to effectively utilize optimization models? The answer is based on key considerations of data, software availability and support staff.
Data. Since the problem of strategic portfolio management is not new, methods are commonly available to adequately satisfy data input needs. Firms have developed their own internal procedures (often with the help of suppliers and management-science consultants) to determine likely values and optimistic-pessimistic ranges for project timing, cost, chance of surviving clinical phases (i.e., risk measures), and NPV. For example, risk measures may be based on internal/industry benchmarks combined with management input (e.g., using "Delphi" methods). With regard to NPV, continual improvement in forecasting accuracy has been realized through ongoing development of new methods and models that can accurately measure the impact of anticipated product positioning and alternative marketing mix strategies.
Software availability. There is definitely a need for user-friendly optimization software. Good commercial and academic modules exist for many optimization methods, but they have to be tied to the specific needs of R&D portfolio management. Minimal criteria for software are that it be PC based, provide a meaningful initial optimal strategy through easy-to-read tables and graphs, and allow efficient interactive evaluation of "what if" adjustments to the initial strategy (known technically as "sensitivity analysis").
Support staff. A project manager with a technical/scientific background plus a quantitatively focused MBA should be adequately prepared to effectively use optimization software specifically developed for R&D portfolio optimization. Most MBA programs sufficiently cover mathematical optimization and statistical methods to allow effective use of optimization software and supporting data input methods. It is important, however, that a project manager be able to interact with and provide feedback to the software supplier’s analysts and consultants for most effective use and customization of software to the firm’s specific needs.
We consider this work to be preliminary. We hope other contributors and corporate R&D insiders can add to our perspective. At a February 2003 Drug Information Association meeting in New Orleans, it was encouraging to note several large enterprise software vendors in attendance. These vendors appear to have found a market in the R&D area. While none of them was presenting optimization features, their data-repository features should prove invaluable in providing the data necessary to drive optimization models.
Pharma is a large industry, and much cost leverage is possible here. The work of earlier optimizers in areas other than R&D (such as Prabha Sinha and Andris Zoltners, who have studied the optimization of sales forces) demonstrates cost improvements in the 10 to 20 percent range. That suggests opportunities for a high level of savings for an industry that clearly need to finds these types of efficiencies going forward.