This post was contributed by Zilliant Pricing Scientist Amir Meimand, who will be presenting at INFORMS in Nashville, Tennessee. In this post, Amir gives a primer on pricing science and explains how machine learning empowers organizations to respond swiftly with price after a change in market dynamics.
Let’s Start at the Beginning: A Primer on Pricing Science
Pricing science started in the early 1980s with the airline industry. These models were mostly based on descriptive and predictive analytics to set price according to forecasted demand and available number of seats. Since about 2000, the application of pricing science has taken off in B2B. Instead of optimizing the offers available in response to very dynamic capacity, these B2B pricing applications enabled optimization based on a particular set of transactional attributes.
At the same time, pricing science methodology evolved to the next stage: employing prescriptive analytics. Prescriptive models have been applied to industry after industry and delivered real, measurable value. Companies that equipped themselves in the early stage earned significant competitive advantages. Not only because of profitability, but also due to business stability and efficiency.
What, Exactly, is a “Prescriptive” Pricing Model?
Although prescriptive models have been used in academia for decades, study after study shows that humans and mathematics working together is far superior to either one alone. By definition, prescriptive pricing models are comprised of three fundamental elements: market segmentation, high-level business objectives and the optimization engine. The latter is the most powerful. Optimization engines connect high-level business objectives with the market price by providing a price recommendation that maximizes revenue and profit based on the relationship between demand and price for each market segment.
By incorporating these three elements, prescriptive pricing models enable companies to consistently determine optimum pricing strategies across all market segments simultaneously. While this is indeed important, it is even more important to be able to continue to determine optimum pricing strategies as market dynamics change over time.
Pricing Strategy, Machine Learning and Market Dynamics
Market dynamics are the factors emerging from market behavior that affect pricing strategies. For example, in the oil industry, the cost of a barrel of crude, total international stockpiles of oil, and the price of alternative energy can each be considered market dynamics. The stability of these factors is one of the most important underlying assumptions of an optimization model. Hence, as these factors change, the model needs to be updated based on the new data.
In some markets, such as the electrical and semiconductor industries, market dynamics change quite rapidly. Thus, updating the model and re-optimizing in real time is not always possible, since the available data are not current and do not reflect the new market dynamic.
In these situations, a machine learning model can be employed to capture the effect of market dynamics on the market price. By combining this model with an optimization engine, the price recommendation can be calibrated based on new market dynamics.
For more information about managing volatile cost conditions, download our in-depth whitepaper, “Pricing Strategies: Managing Prolonged Deflation and Cost Volatility.”
About Amir Meimand
Amir designs and develops pricing recommendation models for B2B businesses in the semiconductor, electrical and rental industry verticals, among others. Prior to Zilliant, Amir’s focus was in B2C, designing promotion-planning software for food retailers. Amir holds a dual doctorate degree in Industrial Engineering and Operations Research from Pennsylvania State University. His doctoral work included applying operations research concepts to dynamic pricing and revenue management. Stay tuned to the Zilliant blog for a recap of Amir’s presentation at INFORMS.