The price updates have thus been able to pass from quarterly cycles in classic retail with monthly or weekly updates which optimize turnover, stock levels and margins. Sometimes called Yield Management, these techniques have proven their worth [4,5,6,7,8] and significantly increase the profits of companies that have invested in these areas. The explosion of online commerce has brought companies into a new era, that of Automated Pricing.

Like financial trading, which has gone from an old-fashioned trading room with orders by phone to automated trading operated by AI (sometimes in nano-seconds), e-commerce is going through a similar revolution. Algorithms of a new kind are being implemented in e-commerce companies. These new techniques exist in the range of pricing management tools because they make it possible to adapt to new update frequencies (daily or even per hour or quarter of an hour when it makes sense). Automated Pricing [9,10] brings new capabilities to manage a much broader and more complex pricing policy. Naturally, we are moving from traditional management of price schedules to real price dashboards.

SKU is the combination of brands / product references / colors / grading / countries of sale / marketplaces / ... When the number of SKUs explodes with hundreds of millions of references to manage, these new automated pricing tools based on engine rules are quickly imposed. The user no longer has the ability to control all the prices due to lack of time and because of the underlying complexity. Human abilities are no longer sufficient to precisely and rigorously update such a large volume of prices, in increasingly short periods of time. This is especially true if one wishes to maintain coherence between all combinations of price segments: products from the same range with different colors or finishes, price difference between cosmetic states in a second-hand pricing policy, ...

With these Automated Pricing tools, the yield manager keeps control of the thresholds that allow PLCs to make optimal decisions at the desired frequency. More and more, the manager no longer performs the final data push on their own, but sets goals and monitors the performance of their strategies. The manager's role is also to set up the safeguards that help the 'control automatons' to verify that the 'decision-making automatons' have not drifted and have indeed respected all the rules and configured thresholds. In the event of alerts, we anticipate backtracking on erroneous decisions.

Price comparators have flourished on the internet as well as marketplaces. On these platforms, rankings are usually performed using the price criterion. This phenomena has contributed to speeding up the implementation of this type of automated pricing tool. The increase in the number of data sources made available in real time via APIs has also been a catalyst for these tools to process in real time from a feedback loop. Indeed, improving the efficiency of these algorithms goes through a machine learning loop that complements simple classic rule engines.

For example, on the launch of a new subsidiary, a new product or a new customer segment, the Bandits Penguin algorithm [11,12] makes it possible to test different pricing approaches within an automaton. Then, it compares their effectiveness to refine the strategy. If several players from the same comparator each use an Automated Pricing algorithm against each other, an anticipatory analysis of the behavior of the opposing engine will be necessary. So as not to make bad decisions regarding the performance of the company, notions of Game Theory should be used. These new pricing tools must indeed make it possible to strike a balance between optimizing turnover and managing the margin [13]..

• [1] - What Is Dynamic Pricing & How Does It Affect Ecommerce? https://www.business.com/articles/what-is-dynamic-pricing-and-how-does-it-affect-ecommerce/
• [2] - Why dynamic pricing is a must for ecommerce retailers ? https://econsultancy.com/why-dynamic-pricing-is-a-must-for-ecommerce-retailers/
• [3] - Fundamentals of Price Management. Hermann SimonMartin Fassnacht - 2018
• [4] - Société Française de Recherche Opérationnelle et Aide à la Décision - www.roadef.org
• [5] - Hotel yield management practices across multiple electronic distribution channels. P O'Connor, J Murphy - Information Technology & Tourism, 2008
• [6] - The hotel yield management with two types of room prices, overbooking and cancellations. T Koide, H Ishii - International journal of production economics, 2005
• [7] - Revenue management and e-commerce. EA Boyd, IC Bilegan - Management science, 2003
• [8] - E-commerce and operations research in airline planning, marketing, and distribution. BC Smith, DP Günther, BV Rao, RM Ratlife - Interfaces, 2001
• [9] - Revenue management through dynamic cross selling in e-commerce retailing. S Netessine, S Savin, W Xiao - Operations Research, 2006
• [10] - Simulating Dynamic Pricing Algorithm Performance in Heterogeneous Markets. HOØ Hilsen - Master Thesis - NTNU - 2016
• [11] - The new frontier of price optimization. D Simchi-Levi - MIT Sloan Management Review, 2017
• [12] - Dynamic online pricing with incomplete information using multi-armed bandit experiments. K Misra, EM Schwartz, J Abernethy - 2018
• [13] - Modèles de bandit : une histoire bayésienne et fréquentiste. Emilie Kaufmann http://chercheurs.lille.inria.fr/ekaufman/Matapli_Kaufmann.pdf - 2017
• [14] - Pricing Lab : Gagner la guerre des prix sans rogner ses marges avec le repricing? https://prestashopday.com/fr/pricing-lab-gagner-la-guerre-des-prix-sans-rogner-ses-marges-avec-le-repricing/ - 2018

Antoine Jeanjean (PhD)