Rules based vs. machine learning based pricing

Rule-based pricing or rule engines are the traditional way of pricing in retail and online retail. Most online retailers change prices using this methodology. In this case prices are changed according to static formulas. This type of pricing can also be dynamic. For example, a retailer can set the rule to always match the third cheapest competitor price. This way prices can change everyday or even multiple times per hour. There are two main shortfalls of rules based pricing:

  • It takes a lot of manual effort to administer the price formulas and
  • price rules are often based on parameters that are easy to measure but entail only limited information about customer behavior

 

Retailers that use rule-based pricing give away profit and revenue potential because their pricing is not based on the willingness-to-pay of their customers.

Next generation dynamic pricing based on machine learning algorithms

On the contrary, advanced machine learning algorithms can measure the willingness-to-pay of customers. To be exact they measure price elasticity which is the key to set optimal prices. Combined with advanced forecasting algorithms they can model the price demand curve of each product. Based on this, companies can steer prices automatically towards their company goals. They do not have to create rules and test their performance. Their pricing reacts dynamically and automatically to a change of customer behavior. As a result, retailers apply discounts more differentiated and smarter.

Real AI vs. fake AI

Customers who source machine learning based pricing software need to be aware that machine learning and AI are buzz words that are often used inflationary. A basic AI system analyzes existing data and uncovers hidden patterns. Based on this, it can make predictions or even prescribes actions. More advanced algorithms use methods called deep learning and reinforcement learning. They outperform basic AI methods by gaining more insights out of existing data. They are also able to process larger amounts of data sources. Most importantly, these methods continuously learn and improve their results.

At 7Learnings we have more than 10 years experience with machine learning technology. We helped some of Europe’s leading online retailers to apply them and use them to optimize prices. If you want to learn more about the potential of next generation dynamic pricing and see it in action, reach out to us and book a product demo.

Today, dynamic pricing is widely used in retail and especially online retail. Its use becomes more common and popular with pricing managers as the driving factors of its application spread to other industries (e.g. B2B sales). Most recently, leading retailers started to use machine learning technology for dynamic pricing with great success. There are many different ways to conduct dynamic pricing. It is important to look out for pitfalls as some companies experienced backlash from customers or were even suspected to violate anti-price discrimination laws. So what are the factors that drive the use of dynamic pricing in retail? And what are the different types of dynamic pricing?

Drivers of dynamic pricing in retail

There are four main factors that drive the significance of dynamic pricing for retail businesses:

  1. Importance of pricing: Product pricing strategy is a crucial aspect for retail companies that directly affects profit and growth. Pricing decisions have direct impact on the main business activities such as marketing and supply chain management. Prices should reflect and complement changes in these business departments. 
  2. Data explosion: Over the last years we saw an explosive growth of data generation and it is likely that the growth rate will increase even more. Especially in retail, customers leave a data trail each time they make a purchase on the web, use their credit or loyalty card. Based on this data, companies can change prices quickly to reflect changed customer behavior. 
  3. Rising complexity & transparency: Selling through online channels has increased price and information transparency dramatically. Changes of external factors (e.g. competitor prices) immediately impact one’s business. To stay competitive companies need to adapt their prices quickly to a changed market environment.
  4. Enabling technologies: Just a few strokes on the keyboard can change prices online. Simple pricing software enables automatic price changes e.g. based on competitor prices. More recent developments in artificial intelligence allow an even higher degree of automation with additional increase in performance.
 

Types by input factors

Forms of dynamic pricing can be differentiated by the main input factors that influence price changes and the technology or main methodology in use. In practice most companies use a combination of different types to tap the full potential of pricing. 

Various factors can trigger price changes. The most commonly used factors are the following:

  • Supply: This is very common in industries with a large share of fixed cost like hotels, airlines or ride sharing. With a decreasing amount of available supply (seats, free rooms) prices regularly increase. In retail, supply plays a major role for stock items especially if they are perishable or cannot be sold after a certain time frame (e.g. fashion products). In this case, retailers often use mark-down pricing where prices are reduced to the end of the season. 
  • Demand: An increased interest for a product can trigger a price increase. Following this strategy, retailers try to capture a larger part of the willingness-to-pay of their customers. A good example is a popular kind of sneaker. In an extreme form, we can sometimes observe price gouging where sellers try to take advantage from demand surges. One example was the increase of prices of face masks during the corona crisis.
  • Competitor prices: Competitor prices are very often used in online retail as one of the main factors to trigger price changes. Especially if competitors sell exactly the same product it is easy for customers to compare prices. Here the difficulty is to determine which products are really price comparison products and at which price difference the customer will change vendor. 
  • Characteristics of customer: This practice of dynamic pricing is often referred to as personalized pricing. Here the price varies for different kinds of customers. The idea is that different customers have a different willingness-to-pay and can therefore be charged different prices. A simple example is a price discount for senior citizens for a museum visit. In online retail personalized pricing is usually put into practice with coupons. Companies that use this form of dynamic pricing need to look out for a potential backlash of their customers that feel that this kind of pricing is unfair.

Types by methodology and technology

Even though some of these factors seem common sense like supply and demand. In practice, it is often very difficult to measure them. Rarely companies act in a perfect test environment where only prices are changed while other demand influencing factors stay the same. Often many things change at the same time (e.g. competitor prices, marketing spend, seasonality). That is where technology and methodology comes into play. Here we can differentiate between two types of dynamic pricing:

  • Rule-based pricing: This is the most commonly used method of dynamic pricing. In this case, a fixed formula determines prices and price changes. One example is  a rule to set a minimum margin of 20% or to directly match prices of competitors. There are two main shortfalls of this methodology: It takes a lot of manual effort to administer the price formulas and the price rules are often based on parameters that are easy to measure but entail only limited information about customer behavior. Often companies only reach suboptimal results using this kind of pricing method.
  • Machine Learning based pricing: Machine learning technology is a huge enabler for price optimization and automation. Machine learning models can measure price elasticities for different price points or even model the whole price demand curve for each product. Next generation software solutions based on this technology run price optimizations towards company goals automatically without the setting of price rules.

7Learnings has helped some of Europe’s leading online retailers to improve their pricing processes with machine learning technology. If you want to learn more about the potential of next generation dynamic pricing and see it in action, reach out to us and book a product demo. 

Product

  • Intro
  • Soultion
  • Demo
  • Glossary

About

  • Company
  • Founders
  • Culture
  • Career

Product

  • Blog
  • News

Follow us

Copyright 2018 © All rights Reserved by 7Learnings.