SoundCommerce’s new product, Reactor, creates a metadata layer to enhance data management and reduce cloud computing costs, aligning with the company’s mission to make data accessible for both humans and machines. In this show, Eric Best, Founder and CEO of SoundCommerce, discusses the challenges retailers face with data cleanliness, the critical role of semantic layers in preparing data for AI, the company’s latest product, and Best’s vision for expanding the company’s AI-driven data capabilities across diverse industries. On discussing the benefits of Reactor, Best says, “We take that low level, call it bronze and silver data prep work out of the cloud data warehouse, thus reducing cloud computing costs for our enterprise customers.” 

Eric Best’s background and the founding of SoundCommerce

• Best discusses his experience leading to the company’s creation. Best shares his vision and background in technology, noting how his work in e-commerce shaped his insights into data-driven retail solutions.
• Best begins by referencing OpenAI’s recent switch to a for-profit model, discussing its unprecedented $150 billion valuation and the power of AI to transform industries. Best emphasizes how AI’s development impacts both everyday life and business strategy across sectors.
• SoundCommerce was founded five years ago to address a specific data issue in retail: handling and making sense of fragmented customer information. Over time, its capabilities evolved to support data handling needs in diverse industries beyond retail.
• Best highlights how SoundCommerce aligns with the rise of cloud data infrastructure giants like Snowflake and Databricks, offering faster, more efficient data solutions. SoundCommerce enables streamlined data onboarding, addressing client needs for flexibility, speed, and cost efficiency in data processing.

Addressing the challenges in retail AI and data cleanliness

• Best talks about the unique challenges that retailers encounter when trying to optimize AI in their operations. Retailers face significant obstacles, particularly in unifying data for advanced analytics, even when it’s stored in modern cloud environments.
• Best underscores that despite easier cloud access, the quality of incoming data remains a top concern. He emphasizes the need for meticulous data preparation, such as cleaning and labeling, to make AI models and analytics more effective and relevant.
• Best introduces the concept of a “semantic layer”, a structured metadata layer critical for defining and organizing data, which is often missing in most data teams. Without it companies struggle to generate accurate, actionable insights from their data.
• Best explains the advanced queries and data models that are essential in retail, such as predicting inventory needs, assessing the long-term value of customers, and measuring the success of advertising efforts.

SoundCommerce’s expansion beyond retail and AI adoption obstacles

• While the company initially focused on retail, its applications now extend to industries like travel, hospitality, and consumer goods.
• Best discusses broader AI adoption challenges, highlighting key barriers such as high costs, privacy concerns, regulatory limitations, and scalability issues. Best identifies these as common hurdles that restrict AI’s potential in many fields.
• Best outlines three core requirements for effective AI adoption: carefully crafted prompts to guide models, robust and trustworthy AI models themselves, and, importantly, well-prepared data. All three are crucial for businesses to get real value from AI.

Why data preparation is a key factor in AI efficiency

• Best delves into the importance of data preparation, pointing out that many AI models rely on data sourced either publicly or through costly licenses. Best stresses that the quality and relevancy of this data play a pivotal role in model success.
• Best gives an example involving The Wall Street Journal, which sold its extensive archive to OpenAI for $100 million, demonstrating the high demand for exclusive, high-quality data sets. Proprietary data of this caliber is often a crucial differentiator.
• Best discusses the common challenges of outdated data used in training which can limit AI effectiveness, so access to timely, relevant data is essential. AI models need continuous updates to align with real-world changes for better accuracy.
• Best introduces us to Reactor, SoundCommerce’s latest product which creates a metadata layer that improves data comprehension. This layer allows both AI models and human analysts to better interpret the data, adding value to analytics processes.

Introduction to Reactor, a SaaS solution for advanced data preparation

• Best explains that Reactor is a SaaS product licensed for a fixed fee, which integrates seamlessly with major data platforms like BigQuery and Snowflake, and will soon support Databricks.
• Reactor ingests raw data from various sources, applies a semantic layer to structure it, and then loads the data into a warehouse for more refined analysis. This process ensures that data is both accessible and optimized for AI-driven analytics.
• Best highlights Reactor’s cost advantages, describing how it streamlines data preparation, which reduces expenses associated with cloud computing. By automating data filtering, companies can avoid unnecessary costs tied to data management.
• Reactor further enhances cost-efficiency by handling what Best calls the “dirty work” of data prep. Its capabilities help reduce expenses for cloud computing by automating many preliminary stages of data processing, offering a valuable resource to companies.

The role of future-proofing data for effective prompt engineering

• Best discusses AI-related costs, such as egress and ingress fees involved in moving data in and out of cloud platforms. Best explains how Reactor’s design logs raw data upstream, allowing clients to selectively transfer data into the warehouse.
• Best underscores that future-proofing is critical as keeping raw data copies ensures that businesses can use or revisit data for future needs as technologies evolve. Best stresses this as an investment in data resilience, enabling businesses to adapt flexibly over time.
• Best also describes the importance of a “data dictionary,” a standardized definition of terms that supports consistency across various uses of data. It is vital for effective prompt engineering, ensuring AI models interpret and generate responses accurately.

The importance of strategic AI adoption and SoundCommerce’s vision

• Best shares his advice on how companies can strategically adopt AI tools. Best suggests that companies should first focus on centralizing their data to allow seamless access to AI models, laying a strong foundation for future AI projects.
• Best explains that SoundCommerce is not competing with cloud giants like AWS, Google, or Snowflake, but instead provides complementary data preparation services that add value to these larger platforms by improving data quality.
• Best shares his background, including his roles at Amazon and in customer data platform development, and managing a consumer brand. Each experience, he says, was instrumental in forming SoundCommerce’s mission to streamline data solutions for modern enterprises.

What are SoundCommerce’s future expansion plans? 

• Best shares SoundCommerce’s plans for the upcoming year, saying that Reactor is currently expanding beyond retail to serve clients in other sectors. This shift reflects a broader vision for the platform.
• Best discusses the need for SoundCommerce to expand its marketing and sales efforts to capture audiences beyond its initial retail focus, highlighting a strategic shift in the company’s target market.
• Finally, Best talks about SoundCommerce’s ongoing work in refining AI models through better data training. Best emphasizes that this process is iterative, with each data improvement enhancing model accuracy over time.

Guest: Eric Best (LinkedIn)
Company: SoundCommerce (Twitter)
Show: Let’s Talk

This summary was written by Emily Nicholls.