How AI taps data to make ecommerce more dynamic

Product pricing plays a critical role for every product, especially in ecommerce. According to Shopify, global ecommerce sales are expected to total $5.7 trillion worldwide in 2022. However, determining the right price for your goods and services can be tricky and requires large volumes of data to be effective: Should you use static prices, monitor those of your rivals or mix the two? The answer lies in setting a pricing strategy that offers flexibility to make changes when needed.

Traditionally, businesses priced their products and services based on gut instinct or by employing external consultants to manually assess cost, supply and demand. However, in the digital era, businesses have access to vast volumes of data that can be used to predict what impact a slight change in price can have on the demand for a product, — while also taking into consideration many external factors like economic conditions, competitor pricing, seasonality, etc.

Such data-driven customization initially originated in recommendation systems, where an algorithm predicts what you might like to buy, pushing to increase a product’s profit margin further. For instance, Amazon’s algorithm predicts what products you are most likely to buy, Netflix suggests movies you’re likely to be interested in and Spotify suggests trending playlists.

Although, recommendations can be somewhat static. A recommendation system alone can only suggest products, with the user choosing to buy them or not. Further optimizations can be done to the architecture through data science to offer more customized services. Dynamic pricing is one such strategy.

A critical tool for modern businesses

Price optimization and revenue management are two of the most significant issues every business needs to address. However, incorporating machine learning (ML) into the mix simplifies both tasks and makes them more efficient. Companies can better understand the present situation of their market, make adjustments as needed, and optimize their possibilities to increase revenue by processing data through dynamic pricing algorithms.

While traditional dynamic pricing algorithms use historical data to estimate the best prices, modern dynamic pricing algorithms leverage more data and artificial intelligence (AI) and ML capabilities to predict market trends better and optimize prices accordingly. Such a method of price optimization allows a company to have real-time pricing adjustments and efficiently respond to the marketplace to organize product campaigns accordingly and reach their goals.

Gopikrishnan Konnanath, SVP and global head of engineering services and blockchain at Infosys, says that artificial intelligence (AI)-based dynamic pricing allows businesses to customize product pricing in accordance with sales and market trends to increase competitiveness, as well as profits and revenue. 

“Businesses can reap multiple benefits by combining AI, machine learning and algorithms to shift away from static pricing and instead use data to determine price points,” Konnanath told VentureBeat. “This helps organizations navigate the rapidly evolving digital economy by responding to changes in real-time and successfully implementing a data-backed pricing strategy.”

Advantages of dynamic pricing

Setting the right price for an item or service is a recurring economic theory dilemma, since many pricing techniques depend on the desired outcome. It also varies from company to company, with one seeking to maximize profitability on each unit sold or the overall market share, while another may want to enter a new market or safeguard an existing one. Different scenarios can coexist in the same organization for various goods or customer segments.

Dynamic pricing providers use various techniques to enhance pricing engines’ effectiveness. For example, many dynamic optimization algorithms use next-gen neural networks capable of processing billions of pricing scenarios, ensuring the integrity of results with a price-effect prediction accuracy of 90–98%. However, developing a forecasting model is a tedious process that varies based on the specific goals and demands of an ecommerce business.

Modern dynamic pricing architectures also analyze real-time data on competitors’ prices and stock collected from websites using web scrapers or robotic process automation (RPA) bots. It evaluates many internal factors, like stock or inventory, KPIs, etc. And also evaluates external factors, including competitor prices and demand, to generate prices that align with a company’s pricing strategy. 

Konnanath believes that aside from increased profit and pricing flexibility, adopting an AI-based dynamic pricing strategy also allows businesses to improve market segmentation, which is especially useful for companies operating across international markets or different target groups. 

“Companies can also manipulate prices to increase sales during slow periods and avoid unsold products. When coupled with promotions optimization, companies can increase the promotional sell-through and manage the inventory more effectively,” said Konnanath. 

This ability of a business to respond to current demand, rationally use its inventory or stock, or develop a brand perception through specific pricing decisions allows it to stay afloat no matter the current market condition. 

Talking about the different use cases for dynamic pricing across industries, Konnanath said that dynamic pricing has taken off across hospitality, ecommerce, and tourism industries most impacted by demand and global changes.

“Even in B2B scenarios, suppliers have applied AI-based dynamic pricing to improve their wallet share in CPG and process industries,” he said. 

Types of dynamic pricing strategies

Two of the most popular approaches for dynamic pricing are the following:

• Rule-based automated system: In this strategy, the pricing algorithm relies on predefined rules and is executed under human supervision. A domain expert defines several “what-if” rules that cover different scenarios so that the model can adjust itself accordingly. This algorithm depends entirely on the past knowledge captured and is not as flexible in responding to unforeseen events.

• Price-optimization system: This approach utilizes self-learning ML models without human intervention. It is best suited for airlines, hospitality, and ecommerce industries, where several variables impact pricing decisions. Such an AI-based approach depends on a vast amount of data to impact variables on the price. As more and more data is fed into the AI system for training the model, it self-learns through reinforcement-based methods and automatically tweaks the system’s performance.

Enhancing recommendation systems with dynamic pricing 

Dynamic pricing algorithms can provide several benefits when combined with traditional recommendation systems. Through granular customer segmentation, businesses can uncover hidden relationships between data points for generating better customer recommendation characteristics, including behavior patterns, and determine customer persona groups with high follow-through accuracy.

Businesses can set up a product to align pricing recommendations with performance metrics of interest — for instance, margin, turnover or profit maximization, inventory optimizations, etc. Using price-elasticity calculation, users can predefine price elasticity to predict whether customers will accept a new price before making a pricing decision. Business rules in such dynamic pricing solutions can be used as additional settings.

According to Dharmesh Mistry, VP of the technology market unit at Capgemini, AI-based dynamic pricing algorithms inculcated with recommendation systems can help companies reduce costs, reduce their carbon footprint (with better logistics management) and, with the right level of personalization, improve customer experience. 

Mistry said that to do so, data must be consolidated across all channels — including competition data and social media feeds, into a data foundation model.

“When analyzing data for such architectures, it should typically pass through demand analysis, a demand predictor (along with demand projection) where the algorithm recommends a dynamic price suggestion based on the business rules,” he told VentureBeat. “Then, this dynamic price for the product can be further pushed into the front-end customer-facing channels.”

Dynamic pricing AI identifies patterns within the data to reveal market pricing gaps and shows missed recommendation opportunities. When trained through multiple cycles of identifying such patterns, the algorithm can be customized further for pricing optimization. For instance, to determine which product should be recommended first and predict an optimal price that proves to be a fit for every customer.

Konnanath said that once businesses understand the full capabilities of recommendation systems, they can seamlessly apply dynamic pricing solutions, which go hand-in-hand.  

“Together, recommendation systems and dynamic pricing solutions can be used to deliver more with the same resources,” Konnanath said. “The two, when combined, can help businesses determine the most cost-effective options for any task and maximize utilization to increase efficiency and ROI.”

Future challenges for dynamic pricing systems

One of the current challenges that AI-based dynamic pricing systems face today is to understand whether a created model will be geographically universal. What is applicable and accurate in the U.S. may yield no effect in other regions of the world due to immense cultural differences between the markets, according to Dmitry Mikhailov, Ph.D., associate professor at the National University of Singapore and the chief scientific officer of Farcana. He added that sourcing and clearing new data for modeling is costly and energy-consuming.

“Every dynamic pricing inherently attempts to rationalize and quantify people’s behavior. However, the problem is that sometimes it is not completely possible,” Mikhailov said. “For instance, people may start gaming the system by forecasting the algorithm behavior; hence, a company’s profits can erode.”

Mikhailov said that it is necessary to consider such a possibility at the planning stage and incorporate defense mechanisms in the algorithm, i.e., including irrational agents in the data pool. 

Similarly, Mistry said that two significant future challenges would be integrating data from edge devices and keeping the physical store’s price changes consistent and in sync with digital channels. 

“Organizations should continue to make investments in sophisticated data and analytics tools. Businesses can also apply pervasive automation for greater personalization and improved customer experience,” he said.