Dear Friends,
In the latest edition of the Game Changer Expert Interview Series, BCG’s Rohan Kadakia explores how companies successfully make the transition from the Cost Game to the Value Game. Rohan identifies the challenges and the key factors that contribute to sustained long-term success in this transition. You can watch the full video here.
In case you missed them, here are some posts since last week’s newsletter.
Reddit ups its ad game: The service is now testing dynamic product advertisements in its feeds. You can read the full post here.
Power in California: Finding the right pricing model for energy in California remains a challenge. You can read the post here.
For this week’s guest contribution, my co-author Arnab Sinha and I asked our colleagues Jacqueline Martinez and Rohan Kadakia to describe how companies with complex portfolios and immense amounts of data are applying GenAI behind the scenes to unlock insights that were previously hard to extract or inaccessible.
How GenAI Makes Messy Data Useful
The buzz around generative artificial intelligence (GenAI) in inescapable, and the retail and industrial distribution sectors are no exception. Much of that buzz surrounds natural language chatbots.
We see the potential for chatbots to drive value by making pricing tools more intuitive. The reality is that many retailers and distributors cannot tap that value unless they resolve some fundamental challenges in their core data and analytics. Using the power of GenAI, we have helped some of them achieve breakthroughs in addressing those challenges by turning messy data sources into useful ones.
Think of the sheer amount of data that any company with a large, complex portfolio collects. While much of that data, such as transaction data, is structured, retailers and distributors also have large sets of unstructured data. For example, e-commerce product descriptions and images are a rich source of unstructured data that traditional, or “pre-GenAI” analyses cannot easily draw on. Finding ways to use that wide range of available data creates the potential for companies playing the Uniform Game (like most retailers), the Choice Game (like major B2B players), and especially the Dynamic Game (like some cutting-edge retailers), to unlock profitable growth opportunities by better understanding customer behavior, product interrelationships, price elasticities, and many other inputs into better pricing decisions.
The key word in that sentence is “potential.”
Much of the potential remains untapped, because the data is unstructured or incompatible, or buried within different systems, online descriptions, untagged images, inconsistent spreadsheets, long product manuals, and complex contracts and purchase agreements.
Prior to the emergence of GenAI, analyzing this data and gaining insights required time-consuming and resource-intensive manual review and data maintenance. The large language models (LLMs) of GenAI, however, make it possible to tap that unstructured data quickly and repeatably. For example, it is possible to extract product attributes from images and descriptions and use those attributes to maintain cleaner internal product relationships (such as having bigger pack sizes offer better value) and to match a company’s products to functional equivalents at competitors to better understand market price levels.
While companies are using GenAI for unstructured data to convert a wide range of messy data sources into usable inputs for pricing models, its application requires a combination of prompt-engineering and data-engineering capabilities designed to support a wide range of business use cases, including pricing. Investing in these capabilities and building momentum with early practical wins can lead to sustainable competitive advantages on the journey to cutting-edge tools. The investments also facilitate more effective decision-making and eliminate some forms of manual work.
The kinds of hidden insights GenAI can derive from messy data
We partnered with an apparel company to achieve an early win by using unstructured data to better understand and optimize promotional effectiveness. This company did not have structured documentation of how promotional pricing was marketed, such as “up-to” versus price point marketing, featured in email, etc. Instead, the company stored this information in unstructured spreadsheets with slightly different language or column orders used for each timeframe and line of business.
To overcome this, we trained an LLM to recognize the data and put it into a usable format. Making this data accessible allowed us to incorporate the drivers of promotional performance in our AI models and provide the company with better promotional recommendations without requiring a major manual data-cleaning effort. We learned that many types of products saw stronger performance from either a percent-off message or a price-point message. Some offers, however, only worked when marketed externally via email, while others worked with or without marketing. Such insights were previously impossible or impractical to unlock systematically.
The biggest wins often come from use cases that unlock value early in a journey to more advanced analytical capabilities. Any retailer or distributor with GenAI capabilities can derive value from turning unstructured data into usable inputs for business decisions. The required investment is generally much lower than investing in large software systems.
Original article can be found here.