Understanding Why Not All GenAI Is Suited for Revenue Growth Management

It’s undeniable that generative AI (GenAI) is transforming industries at an unprecedented pace. It has had a widespread impact everywhere, from Hollywood writers to e-commerce giants, and Revenue Growth Management (RGM) is no exception.

With the right Artificial Intelligence (AI) tools, an RGM department can make significant progress towards achieving RGM maturity. But one of the crucial things to understand when evaluating different AI powered tools is that not all AI is the same. 

In this post, we will look into the different types of AI available to Revenue Growth Managers, the differences in their capabilities, and further understand which is best suited to predict the impact of RGM decisions on KPIs (e.g., revenue & profit) as well as to recommend the optimal course of action for maximizing them.

What Are The Different Types of GenAI?

Let’s start with a definition by an expert on the topic. ChatGPT defines GenAI as:

 “Generative AI refers to a class of artificial intelligence systems designed to generate new content, data, or information that is similar to, but not directly copied from, existing examples. These systems use algorithms and models to create original outputs by learning patterns and structures from the input data they were trained on.”

The most prominent current applications are large language models (LLMs) that are trained on large amounts of texts. These models use GPTs (Generative Pre-trained Transformers) – as in ChatGPT. They can generate human-like texts based on the input they receive, making them capable of tasks such as text completion, translation, summarization, and creative writing. 

Very simplified, these models build their output texts step-by-step by predicting the next word in a sentence based on the previous text (Figure 1). This example highlights, that answers are always probabilistic. The sentence “the price is…” can continue with “high”, “low”, or many other words. So, if you ask the same question multiple times, you will likely get a (slightly) different answer each time.

Figure 1

Other GenAI programs that work in the same way can create images or music by predicting the next pixel in an image or note in a song.

Different types of GenAI models exist to do these tasks. Below is a list of some of the most common types, along with a description of how they function, provided by ChatGPT.

• Diffusion Models
  Diffusion models are a type of generative AI that work by gradually transforming random noise into structured data (like images or text). They do this through a step-by-step denoising process, where an AI model learns to reverse the process of adding noise to real data. These models have been particularly successful in generating high-quality images (e.g., DALL·E, Stable Diffusion) and are known for their ability to create realistic and diverse outputs.

• Generative Adversarial Networks (GANs)
  GANs consist of two neural networks—the generator and the discriminator—that compete against each other. The generator creates synthetic data (such as images), while the discriminator evaluates whether the data is real or fake. Over time, the generator improves by learning to fool the discriminator, resulting in increasingly realistic outputs. GANs have been widely used in image generation, deepfake technology, and enhancing low-resolution images.

• Neural Radiance Fields (NeRFs)
  NeRFs are a type of AI model used to create highly realistic 3D scene reconstructions from 2D images. They work by learning how light interacts with objects in a scene and using this information to generate a continuous 3D representation. NeRFs are particularly useful in applications like virtual reality, 3D modeling, and realistic scene rendering in movies and video games.

Hybrids of these different GenAI technologies also exist, which combine different parts of the techniques named above. Many of the specialized GenAI tools that exist are made up of one of, or a combination of these types of GenAI models.

What Type of GenAI is Right for RGM?

An RGM team may assume they can simply input their data into ChatGPT or Microsoft Copilot, ask “How do I best increase my net revenue by 4%?” and receive the price, promotion, and PPA changes required for this outcome. However, this is not the case, as not every AI tool is the same.

A GenAI LLM such as ChatGPT can speed up certain RGM tasks and increase efficiency to a certain point, but what it lacks is predictive accuracy and the ability to deep dive and optimize all RGM levers. GenAI LLMs like ChatGPT and Copilot excel at tasks like summarizing information and automating communications, but they are not specifically trained to optimize pricing or other RGM levers - the halo effect can therefore lead to wrong assumptions about GenAI’s capabilites. 

Figure 2

Using AI to navigate RGM decisions requires structured, predictive AI rather than purely language-based models. Companies that rely solely on GenAI for RGM risk making inaccurate predictions and suboptimal business decisions. This is due in part, as mentioned, to the fact these tools are not custom-made for the job. Add to this, the fact that some LLM models can ‘hallucinate’ and provide false answers, and it is clear that businesses should avoid using unsuitable AI for Revenue Growth Management decisions.

It is worth noting that even the more specialized pricing and RGM tools do not all leverage GenAI in the same way. Many RGM AI tools simply combine the traditional RGM tools, like price elasticities, with AI suggestions in the same way LLMs like ChatGPT do. This means that while their suggestions are based on real data, and are fast and efficient, they might lack accuracy due to the underlying static price elasticities. This is extremely detrimental to long-term growth.

Whether you are considering building a specialized RGM platform powered by AI or looking into RGM solution providers, it is important to take into consideration the following: 

• Data integration: RGM relies on a very broad range of relevant data sources that are used for decision making. Making these different data sources useful requires several steps including data cleaning (e.g., identification of errors or outliers) or harmonization of sales data coming from different sources that need to be mastered before the data can be analyzed. Also, different forms of data – for example, sell-out and survey data – need to be aligned.

• Sales forecasting: RGM GenAI for sales forecasting should aim to replicate human-like behavior as closely as possible to predict shoppers' subsequent product purchases after different RGM levers are pulled - and not utilize a model that predicts the next word to understand the next purchase. 

• Offer optimization: There is a role to play for GenAI in identifying and recommending the right offer changes that optimize the organization's KPIs, while addressing the challenges of complexity, constraints, and definition of the objective function. This is achieved with Buynomics Decision Guide.

• Recommendation and management of offer changes: It is most important to recognize that such an interface can only be as good as the solution that it interacts with, as the LLM cannot perform key tasks on its own. Therefore, the LLM mainly serves as an interface between the user and the GenAI that can predict how sales react to offer changes and identify optimal offer changes from that. For the AI to provide a useful answer, it must be able to identify which offer changes yield a 5% net revenue increase and of those, which will be considered “best” by leadership. Therefore, a prompt-based interface must satisfy the following requirements:
  • Precisely understand technical terminology of the domain.
  • Assess which options are considered “better” by an organization.
  • Avoid hallucinations.

This requires a structured, multi-layered system. Such systems need to have access to specifically designed and trained models for the key tasks of data integration, sales forecasting, and offer optimization. For an RGM AI tool, it is more important to specifically focus on understanding the key terminology of the RGM domain than being able to rephrase a question about net revenue optimization as ChatGPT could.

Figure 3

Figure 3 shows the overview of how the four GenAI layers for RGM are stacked on top of each other. 

For more information on the key layers that need to be considered for a GenAI RGM tool, download our Generative AI for Revenue Growth Management whitepaper.

With Buynomics Virtual Shopper AI, we combine these layers to create a fully comprehensive model. The Virtual Shoppers AI creates 100,000s or millions of virtual shoppers, each equipped with an individual set of preferences for different brands, product sizes, and other product attributes – and behavioral features, meaning they make purchasing decisions just like real shoppers.

This holistic, AI-backed approach that understands the links between critical revenue levers not only improves the efficiency of decision-making but improves accuracy too.

What Can be Achieved When Using the Right AI Tools for RGM?

As mentioned above, using the right AI-powered RGM solutions unlocks a huge number of benefits for businesses, including the ability to move from a reactive to a proactive RGM strategy.

Want to know where your RGM Maturity currently stands? Take our RGM Maturity survey to find out.

The most efficient way to navigate volatile markets, competitor moves, and inflation is by being able to accurately predict what your shopper will do when standing in front of the supermarket shelf. Without an advanced RGM platform like Buynomics, this is incredibly challenging to predict. 

The right tools enable forward-looking decision-making and foster a data-driven culture within your RGM department, and across the whole organization. Using this cutting-edge ability allows businesses to optimize their RGM decisions by anticipating portfolio, category, market change, and competitor moves. 

This not only benefits the business itself but can be used to reach a triple win that will benefit manufacturers, retailers, and shoppers alike. In turn, proving to retailers that decisions are data-backed, with proof points to show and explain strategic decisions, gives businesses a much better basis for negotiation that will lead to improved relationships.

Take the Next Step To Integrate the Right RGM AI Tool

If you’re ready to reap the benefits of using the right AI tool for your business, Buynomics is here to help.

Buynomics Virtual Shoppers is a custom-built, AI RGM technology that boosts gross profits by 2%- 4% and reduces decision-making time by up to 80% and is trusted by global companies like Danone, General Mills, Unilever, L’Oreal, and Nestlé.

Request a demo today to see how Buynomics can support your RGM team and help you make data-driven decisions!