Modeling

The Farsight Process

For a marketer with vision, anticipating the future is every bit as important as understanding the current environment. Our Farsight suite of forecasting techniques and service offerings use survey data collected from current and prospective customers, as well as historical customer and market information, to develop demand forecasts under varying market conditions and business or product scenarios.  The models are calibrated to properly capture current market conditions and temporal dynamics, and they have been validated to ensure that the projections are trustworthy.

When developing new or reconfigured products and services, conjoint analysis (or "choice-based" analysis) is often used as a platform for modeling customer choices.  The power of conjoint analysis lies in the latitude it affords us to incorporate a relatively large number of product variables or market scenarios into predictive models, while also measuring the importance (and desirability) of various product/service attributes in accounting for market outcomes.  By having customers evaluate a systematic subset of all possible product/service permutations, it is possible to: (1) simulate interest in product/service profiles not explicitly tested; and (2) identify the decision dynamics that account for product/service selection.  Bayesian analytic techniques are typically used to enhance the accuracy of inferences that are drawn from the underlying data.  The result is a data set of customer choice (or preference) input from which predictive algorithms can be constructed to model potential market penetration or share predictions.

In instances where potential new product configuration options are limited (or where decisions about configuration have already been made), relatively straightforward methods for forecasting demand are used, often involving surveys with direct questioning techniques regarding purchase likelihood and potential quantities.  Demand forecasts generated from such survey results, are calibrated using a set of algorithms to reflect anticipated market awareness levels, estimates of overstatement, and information on adoption rates for similar products.

Farsight forecasting models provide the backbone of our market response simulators and have been used by clients in a wide range of applications including:

  • Managing product portfolios to maximize penetration and loyalty across market segments while minimizing cannibalization
  • Assessing product portfolio management decisions
  • Projecting implications of pricing shifts on sales volumes and profitability
  • Forecasting the effect of promotional incentives on short-term and long-term sales, purchase cycles, and net profits
  • Forecasting the adoption of new products over multi-year time horizons
  • Forecasting uptake in different segments
  • Forecasting the revenue benefits associated with R&D investments and clinical research strategies

Conjoint Analysis and Other Choice Modeling Techniques

One of the most widely used advanced choice modeling techniques, conjoint analysis enables us to project market response to a broad range of scenarios or product configurations and to draw inferences about the motivational drivers of customer demand. Conjoint analysis lends itself to many different kinds of raw data projections including discrete choice, rating scales, volume projections and share allocations. Different market situations and product decisions may dictate one approach over another, and surveys must be carefully designed to mimic a "naturalistic" decision framework. Because we know that one size or structure never fits all problems, National Analysts Worldwide can customize predictive algorithms to accept any kind of input data.

Though widely used and exceedingly powerful, conjoint analysis requires skilled application to make key design decisions:

  • Properly selecting, defining, and operationalizing input variables
  • Framing the customer decision appropriately and realistically
  • Making provision for statistical interactions
  • Providing respondents with a "manageable" task to avoid fatigue and data degradation

National Analysts Worldwide employs several advanced approaches to wring maximum insight from conjoint exercises while safeguarding – and even enhancing – validity.  These include:

  • Application of Hierarchical Bayesian (HB) modeling techniques to support more accurate conjoint analysis results while alleviating respondent burden
  • Use of new, proprietary adaptive conjoint models that allow us to explore a substantially larger number of variables than traditional conjoint approaches allow

The resulting data drive the gears of sophisticated market response simulators, enabling clients to reliably estimate marketing response to product configuration and pricing decisions.

For more information on National Analysts Worldwide, please e-mail us or call (215) 496-6800.