Simulation Forecast Models Using @Risk A focus on Market Models for Product Forecasts Copyright 2007 Applied Quantitative Sciences, Inc. About AQS • Applied Quantitative Sciences, Inc. is a service organization that specializes in assisting clients understand and navigate the process of making important decisions in the context of risk and uncertainty. • Our Healthcare Division has as its clients global market leaders in: – – – – Pharmaceuticals Biotechnology Medical device Healthcare delivery systems • Corporate office located in Broward County, FL AQS Healthcare Division Clients Agenda • Why forecast? • What is simulation forecasting and how is it different from traditional forecasting methods • Basic truths about forecasting • Roles and responsibilities within the forecasting team • Components of a representative forecast • Managing validity, reliability and credibility of the forecast • Pitfalls to avoid when reporting outputs • Tools for simplifying the task of simulation forecasting 4 A lot to cover.... • I’ll try to be both informative and brief 5 Why Forecast? • You simply cannot plan without it • Virtually every function in a company must plan – – – – – – Business development Sales Finance R&D Manufacturing and production Supply chain 6 Just because you have a forecast... 7 Doesn’t make it useful 8 Traditional Method • Characteristics: – Inputs are single point estimates – The model relates the inputs mathematically – Outputs are deterministic and represent a single scenario • Can answer questions such as “If things happen the way we hope, we can expect…” Monte Carlo Method • Characteristics: – Inputs are distributions of uncertainty – The model relates the inputs mathematically – Outputs are representative distributions of what is possible, with likelihoods able to be calculated for any value • Can answer questions such as “What is the likelihood of achieving…” Why Use Simulation Forecasting 11 “Any realistic model of a real-world phenomena must take into account the possibility of randomness.” --- Sheldon M. Ross 12 Some Truths About Forecasting • • • • • No-one can predict the future The myth of accuracy You cannot effectively plan without it Few do it well In many large corporations, various functions use different forecasts for presumably the same outcomes 13 No-one can predict the future: forecasting is not fortune telling 14 When one tries to pretend it is, they look more like this... 15 And, when actuals come in... 16 Accuracy vs. Representativeness • Accuracy refers to how well a model predicts actuals – Within sample (to establish the model) – out of sample (to validate model) • Accuracy is irrelevant when discussing simulation forecasts – A well designed simulation model will incorporate all potential futures, and is therefore always “accurate” – The precision of the forecast is determined by the level of uncertainty surrounding inputs – The purpose of a simulation forecast is to provide insight into the range of potential futures and inform decisions based on the likelihood of achieving any level within that range – Simulation embraces uncertainty 17 However, validity and reliability is critical • Predicated on – – – – A sound, reproducible forecasting process Appropriate model specification Appropriate assumption sources Testing 18 The AQS Modeling Process 19 AQS Modeling Standards • • • • • • • • Transparency – Assumptions – Structure Modularity – Increases structural flexibility – Models are more maintainable – Lower total cost of ownership Documentation – All assumptions are documented for source, rationale and date Scaleability Adaptivity to evolving business requirements Portability of results – Designed to provide specific decision support with explicit levels of confidence – Able to integrate with existing enterprise applications Quality Assurance (Every model is reviewed by an independent modeler, with signoff / certification by the individual performing QA) – Model structure – Assumption documentation – Code integrity (every line of code is reviewed for mathematical and referential integrity) Version Control and Traceability 20 Roles and Responsibilities of the Forecasting Team • Sponsor: The person who defines the scope of the project and who will be the primary consumer of the outputs. Responsible for engaging the appropriate resources to establish a valid and reliable forecast relevant to the business domain(s) of interest. • Forecaster: Takes a leadership role in project management, identifying appropriate sources of inputs, and managing potential conflicts of interest or bias • Modeler: Translates information about market dynamics (whether verbal or otherwise) into mathematical representations that produce consistent with what one would expect in the real world 21 Often, more than one role is assumed by a single individual 22 Components of a Representative Forecast • • • • • • • • Segmentation Key events Target population Exclusions Target pool (Target population - Exclusions) Technology adoption Market share Average Selling Price 23 A Model Framework Revenue = Target Pool * Adoption * Market Share * Price Disruptive Risk Instability in population Uncertainty Disruptive Risk Emerging technologies Healthcare economics Study data Correlations Disruptive Risk Disruptive Risk Time of entry Relative advantage Contracting & Distribution Relative sales force size/tenure Study data Disruption to COGS Competitive pressure The fascinating impressiveness of rigorous mathematical analysis, with its atmosphere of precision and elegance, should not blind us to the defects of the premises that condition the whole process. There is perhaps no beguilement more insidious and dangerous than an elaborate and elegant mathematical process built upon unfortified premises. - T.C. Chamberlain Forecast Inputs • • • • • In many strategic forecasts, there are no data from which to extrapolate Garbage in; garbage out Identify who should (and who should not) own inputs Perform formal assumption elicitation interviews Revisit as new data / information are available 26 Incorporating Risk and Uncertainty • None of us know the future • Many of the variables of interest for the model represent uncertain future events. • Someone, whether implicitly or explicitly, will make assumptions and decisions based on these uncertain quantities. • It is preferable to have the most qualified individual make explicit the assumptions that drive the model, so that they may be documented, understood, discussed and (when more current knowledge is acquired) updated. Even a good forecast may not be credible • Simulation forecast models can incorporate a significant amount of complexity • No matter how complex the model becomes, consumers will not trust a “black box” • It is critically important to communicate modeled dynamics and outputs in an effective, actionable fashion – Making a good bet – Risk analysis and mitigation 28 Without Making Their Heads Explode Benefits Beyond the Numbers • Assumption transparency, traceability and updateability • Explicit probability revealed for continuum of potential values • Consistency • Rigor • Credibility • Reproducibility • Provides a framework for valid and reliable management of opportunity portfolio 30 Why is this not routinely done today? • It is hard work – Perceived lack of tools to simplify • No common language • The desire for “the number” (elective ignorance of uncertainty) • Corporate innumeracy 31 Commercial Simulation Engines Make Probabilistic Long-Term Forecasting Accessible to Any Organization 32 Sophisticated Business Models Made Easy • A business-modeling tool that dramatically simplifies the task of building sophisticated, complex business models, with or without Monte Carlo Simulation. AQS Model Builder instantly creates model components for: – – – – – – – – – – • Technology adoption Market share Probability and timing of product launch Probability and timing of key market events Disruption Dynamic transitions from one value to another Dynamic transitions by a percentage of the base value Replacement/Replenishment model Dynamic pricing models And MORE! Cut the time you spend building models to a FRACTION of what it takes today. Decide With Confidence Forecasting Simulation Optimization Statistical Analysis •Portfolio Management •Strategic Planning •Supply Chain Management •Resource Planning •Pharmacoeconomic Analysis •Risk Management Applied Quantitative Sciences, Inc. 649 San Remo Drive Weston, FL 33326 561-433-8206 www.aqs-us.com Thank you.