Methods for automatic planning and price optimization are becoming increasingly more important in retailing. They are among the most innovative of analytical methods. prudsys AG is recognised as a pioneer in this field and in recent years has developed and successfully implemented a comprehensive modular system for dynamic planning and price optimization. The methods are based primarily on regression analysis for estimating price elasticity (sparse grids) combined with reinforcement learning.
The ability to track changes in sales trends in real-time is of central importance to dynamic planning because it enables an immediate response to changing market conditions and also enables the optimization of storage. In the first instance the algorithms use transaction data, i.e. sales figures, and then the master data for a product. Special off-line procedures can then be used to prepare forecasts about new products.
Methods for dynamic price optimization offer more than just forecasts. They will also deliver optimised prices in real time. The reinforcement learning framework plays a role here. Existing algorithms for automatic price optimization have the drawback that they vary price too little and do not therefore provide sufficient empirical data on which to base forecasts. Dynamic price optimisation fills this gap. The continuous and systematic changes of prices systematically fills the gaps in the data and thereby enables a more precise estimate of price elasticity. The framework for dynamic price optimisation is very wide ranging and enables the definition of: optimization parameters such as sales and earnings; strategies such as high price/low price, demand/competition oriented, anonymized/personalised; constraints such as price limits, time frame, variance, degradability.
Advantages
Rapid response and matching to changes in customer buying habits
Combination of off-line and online learning.
Planning: Sales forecasts for new products
Price optimisation Automatic price setting
Price optimisation Optimisation of target parameters, strategies and complex constraints
Simulation function for testing different prices and strategies
Integration
The planning and price optimisation procedures are packaged and incorporated in the XELOPES library. There are also off-line and online variants of the algorithms; online algorithms play a major role. Complex models can be serialised using special PMML add-ons. The prudsys RDE is the central application in the planning and price optimization procedures. The RDE server acts as the administrator with a graphic interface featuring a comprehensive range of administration and processing functions.
