Visualization and Interactive Filtering of Strategy Driven Portfolios as an Alternative to LP Optimization

摘要

Linear Programming approaches to Portfolio Optimizationrnrequire much effort to “teach the computer” the PetroleumrnBusiness. We must described to the optimization program therncosts, potential rewards, and uncertainties of every investmentrnopportunity. For an optimizer to generate realistic,rnexecutable, investment programs, we must also encodernconstraints that define the strategic and geologic dependenciesrnbetween investments, the timing flexibility that exists withinrnand between each opportunity, the working interests that existrnor can be negotiated, and their competition for scarcernresources such as rigs, people, and partners’ approvals.rnStrategy gets encoded as portfolio goals to be met.rnPerhaps we can employ this effort more effectively. Wernought not search for one magical optimum portfolio dependentrnupon one strategy masquerading as thousands of LPrnconstraints.rnThis paper describes a portfolio analysis process thatrnquickly creates near optimum executable portfolios on each ofrnthousands of strategies and portfolio value criteria. Eachrnportfolio and its measures, including confidence curves,rnaccumulate in a database. We use laptop-capablernvisualization software to display and compare the portfoliosrnon dozens of performance measures. Using interactivernfiltering we can apply portfolio goals and resource constraintsrnto the cloud of portfolios quickly narrowing down to a fewrnportfolios that meet all our criteria. Finally, from these bestrnportfolios we learn which opportunities are funded in all,rnmost, some, or none of the final portfolios. In the end, this isrnthe information we seek: what are the smart investmentrndecisions to make, not which one portfolio we should execute.rnBusinesses who adopt this process benefit several ways:rn1. Using interactive filtering to apply constraints means yourncan change resource levels very late in the process. 2. Yourncan better set goals and negotiate resources after seeingrnbeneficial potential trade-offs between portfolio measures.rn3. You define your constraints using the measures of totalrnportfolio performance you save to the database, including nonlinearrnmeasures from uncertainty distributions. This way isrneasy to apply constraints such as “P90 of NPV >= 500” orrn“Number of Funded Projects <= 30”; constraints which arernimpossible with linear programming.