Why Neural Networks Fall Short in Price Optimization

“The value of AI isn’t in the systems themselves. Rather, it’s in how companies use these systems to assist humans—and their ability to explain to shareholders and the public what these systems do—in a way that builds trust and confidence.” - What is AI (artificial intelligence)?, McKinsey.com 

There’s no doubt about it, we’re in the middle of an era that will be shaped by AI and AI-driven technologies. ChatGPT opened up the door for everyday AI use and quickly became one of the fastest-growing applications in history. The global market for AI is growing at an exceptional rate and, according to Forbes, is expected to reach $15.7 trillion by 2030, more than the current output of China and India combined.  

But when it comes to integrating AI into business and software solutions, our philosophy is to not just show off new technology. AI must be pragmatically applied to solve specific pain points and business challenges. We use AI judiciously when and where needed rather than simply leveraging the technology because it’s an industry buzzword. 

That brings us to neural networks – the latest AI/ML trend that some companies have taken to use in their pricing technology. In this blog post, we’ll break down what a neural network is, and why, although beneficial, it’s not the best fit for price optimization models. 

What is a Neural Network?

Neural networks, also known as artificial neural networks, are a type of machine learning that uses large amounts of historical data to train models that are built to mimic the human brain. Neural networks are good for grouping data and recognizing patterns, speech, and images (as opposed to generative AI which generates net new images and text). 

According to AWS, this is how neural networks work: 

The human brain is the inspiration behind neural network architecture. Human brain cells, called neurons, form a complex, highly interconnected network and send electrical signals to each other to help humans process information. Similarly, an artificial neural network is made of artificial neurons that work together to solve a problem. Artificial neurons are software modules, called nodes, and artificial neural networks are software programs or algorithms that, at their core, use computing systems to solve mathematical calculations. 

Today, neural networks are used across several industries, including for medical diagnosis in healthcare, targeted marketing in social media, forecasting and historical data processing in financial institutions, facial recognition in surveillance, and much more.   

So, while there are many extremely beneficial and practical uses for neural networks in the real world, there are limitations, including a glaring one concerning data.  

From The-Scientist.com: 

Because [neural] networks learn from the data they are trained on, they learn the patterns represented in only that dataset. If the test dataset overrepresents or excludes possible examples, such as only seeing beaver dams built on rivers, then this will bias the network to be less likely to recognize other possible data points in a group, such as a beaver dam built on a lake. 

Understanding the Drawbacks of Neural Networks for Price Optimization

Implementing neural networks for price optimization in a B2B setting may be difficult and problematic because of the substantial data requirements and the lack of flexibility or controls these models bring.   

Let’s further examine these drawbacks: 

• Substantial Data Requirements: Neural networks require large amounts of historical data that can be difficult to obtain for most manufacturing and distribution companies. Additionally, companies need to ensure that their data is clean and ready to use as bad or skewed data could impact results. 
• Data Sensitivity: If there was a heavily discounted price point in the past, that same data is used to train the neural network, which influences future prices. This could unintentionally lower prices over time.  
• Lack of Transparency: Teams are unable to interact with neural networks, so there is little insight into how or why the neural network recommended a price, which will likely hinder the operationalization of a desired pricing strategy. 
• Complex Implementation: Implementing neural networks is more expensive as you’ll need a data scientist to work on the model implementation. Plus, the process will take longer if you want to include additional data points such as competitor data, market data, and additional data sources not usually used for segmentation.  
• Lack of Flexibility: Neural networks treat customers individually without grouping them into categories like small, medium, and large. This compromises the ability to leverage strategies based on characteristics like customer size which could impact the alignment of prices for different sized customers.  

Take Strategic Control Over Pricing with Zilliant Price Optimization

Zilliant is actively working on making it easier, not harder, for manufacturing and distribution companies to leverage price optimization. This means simplifying our implementation process, making the solution more transparent, and decreasing the time to value. 

Here are some key benefits of our market-leading price optimization solution: 

• Highly transparent and user-friendly with segmentation models that allow you to have close interactions with your model and apply complex pricing strategies. 
• The ability to leverage categorical data to match similar customers with similar attributes helps improve the accuracy of your segments with less historical data. 
• Predictive Pricing and scenario modeling help users understand the revenue and margin impacts of pricing strategies before executing them, which is key to driving financial benefit. 
• Access to Zilliant’s team of data scientists for education on segmentation models to ensure you have a framework to understand pricing decisions. 
• Users can instruct the model to adjust and apply higher price points for those segments if historical data reveals that certain segments were consistently priced too low. 
• Our widely recognized constraint-based price optimization approach offers the flexibility to implement pricing strategies tailored to meet a customer’s unique business needs. 

Ready to reap the benefits of price optimization? Let’s chat.