In late 2010, Supply Chain Matters published a commentary describing Lokad A Twist in Cloud Computing- Forecasting Mathematicians On-Demand.  This rather unique cloud based technology provider provides its customers with highly sophisticated product demand forecasting services which differentiates itself on the sophistication of its staff of mathematicians who take on challenges reflected in difficult or complex forecasting problems. We sometimes refer to the company as “mathematicians on demand’.

CEO Joannes Vermorel recently corresponded with us regarding his team’s current efforts in the area of quantile forecasts, which represent the most significant upgrade for Lokad technology since the company’s launch.  As opposed to deterministic or mean-driven forecasts where respective forecast weighting are averaged, quantile forecasts introduce a purposeful bias in the forecasting algorithm. Quantile methods can be viewed as a stochastic method for forecasting.  This method can be routinely used in retail, distribution-driven or process manufacturing environments where “spike” demand is prevalent. It can be a more accurate method to address seasonality of demand such as the manufacturing of seasonal products or goods. As Vermorel describes, rather than forecasting deterministic average behavior, quantile forecasting embeds specific distribution uncertainty for each forecasting period (intra-year, intra-month, intra-week). A quantile approach assumes that a perfect accurate forecast cannot be achieved because many variables are not clearly defined, and instead, attempts to balance the risk of over or under forecasting.

Lokad reports that it continues to test its quantile methods on many industry verticals including the production of auto parts, electrical supplies, textile products, spare parts and packaging materials. Keep in mind that the company generally takes on rather difficult or troublesome forecast challenges to begin with. Vermorel and his team are also rather pragmatic in the view that the goal is not to have the most accurate forecast, but rather a more automated means to determine the best best response to fulfilling product demand under challenging constraints.

Bob Ferrari