Linear Programming approaches to Portfolio Optimization require much effort to "teach the computer" the Petroleum Business. We must described to the optimization program the costs, potential rewards, and uncertainties of every investment opportunity. For an optimizer to generate realistic, executable, investment programs, we must also encode constraints that define the strategic and geologic dependencies between investments, the timing flexibility that exists within and between each opportunity, the working interests that exist or can be negotiated, and their competition for scarce resources such as rigs, people, and partners’ approvals. Strategy gets encoded as portfolio goals to be met.

Perhaps we can employ this effort more effectively. We ought not search for one magical optimum portfolio dependent upon one strategy masquerading as thousands of LP constraints.

This paper describes a portfolio analysis process that quickly creates near optimum executable portfolios on each of thousands of strategies and portfolio value criteria. Each portfolio and its measures, including confidence curves, accumulate in a database. We use laptop-capable visualization software to display and compare the portfolios on dozens of performance measures. Using interactive filtering we can apply portfolio goals and resource constraints to the cloud of portfolios quickly narrowing down to a few portfolios that meet all our criteria. Finally, from these best portfolios we learn which opportunities are funded in all, most, some, or none of the final portfolios. In the end, this is the information we seek: what are the smart investment decisions to make, not which one portfolio we should execute.

Businesses who adopt this process benefit several ways: 1. Using interactive filtering to apply constraints means you can change resource levels very late in the process. 2. You can better set goals and negotiate resources after seeing beneficial potential trade-offs between portfolio measures. 3. You define your constraints using the measures of total portfolio performance you save to the database, including nonlinear measures from uncertainty distributions. This way is easy to apply constraints such as "P90 of NPV >= 500" or "Number of Funded Projects <= 30"; constraints which are impossible with linear programming.

You can access this article if you purchase or spend a download.