Grocery Carts to Data Streams — How Kroger Harnesses Data to Enable Customer Delight

The Kroger Co. is one of the largest retailers in the U.S. As per the company’s official LinkedIn page, the Kroger umbrella has “half a million associates across 2,800 stores in 35 states, operating two dozen grocery retail brands and 34 manufacturing and 44 distribution locations.”

In conversation with Denodo Enterprise Data Account Executive Dave Nixon, Todd James, Chief Data and Technology Officer at 84.51˚, (Kroger’s retail data science, insights, and media company), dives deep into the company's data-centric strategy, exploring how it leverages data to personalize customer experience, strategizes for AI, and empowers its workforce.

Speaking about the data strategy for enabling customer experience and customer 360, James highlights three key areas.

1. Data should be considered within its contextual frame. “The idea of domain-specific architecture and strategy is very important. You want to provide as much ownership and flexibility at the edge within the domain so that you can have the right context around the data itself.”

2. Taking steps to improve data quality, safeguard it, and enrich it as an asset across all aspects.

3. There should be a beneficial value exchange. “We want to leverage data in a way that returns benefit and value to the customer.”

James goes on to reveal that Kroger has made it significantly more convenient to fill online shopping carts making the journey ~4.5x faster over the last few years through the use of data for various initiatives like product recommendations, email outreach, etc.

Enabling customer delight through data-driven personalization

Kroger serves roughly 23 million digitally-engaged households initiating about 500 billion “start my cart” recommendations every year. James says that data acts as the foundation, which is managed through automation and relevancy sciences to offer a relevant and personalized customer experience in the purchase journey.

He tells Nixon that a big focus area for the organization is to continue improving and augmenting the personalization science, the data, and the experience itself in a way that drives value and simplicity for customers – right from the time they interact with a touchpoint like a promotional email to when they start their cart.

James further adds that Kroger is incorporating semantic search capabilities to help customers find relevant products or a relevant substitute if something is out of stock. He further mentions leveraging relevancy sciences for aspects like “Did you forget something?” prompts as customers near the end of their purchase journey.

“It's all about the customer. We're trying to make sure that there is a relevant personalized experience at every aspect of the customer journey, through their lens, in a way that makes their life and shopping easier, and brings more value into their experience,” James adds.

Balancing innovation with responsible data practices

James highlights that ethical and data privacy reviews are embedded throughout the organizational use of data. “We are willing to slow down the process of development to make sure that we are taking appropriate steps to safeguard data. Our regulatory environment management is consistent with the ethics we value both as individuals and as a company.”

Speaking about the nuances of risk and compliance around data, James says that while the business owns the data, it's managed by technology teams familiar with the domain. He elaborates that data mesh allows for data segregation based on various factors, including regulations.

While data owners (the business) have a say in how their data is used and invested in, data management remains with the tech team, who have the necessary context and understanding of the data for both operational as well as analytical uses.

James further emphasizes the challenge of managing operational versus analytical use cases. While operational tasks are generally fueled by domain-specific data, analytical ones often involve complex cross-domain data interactions akin to interstate travel. He advocates for a balanced approach providing autonomy at the edge while maintaining a thin layer of governance at the center.

On a similar note, James mentions working with other leaders in the technology organization for domain-based strategies and standards to ensure data usability across the enterprise. He stresses the importance of collaboration and standardization to produce usable data products for all stakeholders.

Strategizing for AI

Kroger has a heritage of experience in areas like advanced analytics and artificial intelligence techniques backed by its analytics arm 84.51°. James says that over the past 18 to 24 months, AI technologies like generative AI are increasingly becoming more democratized and integrated into many vendor-based platforms and products. While this opens up new possibilities, the risks around data haven't diminished. “Because it's easier, doesn't make it less risky.”

James adds that companies that have been investing in the right processes, controls, and systems, are in a prime position to be able to manage the democratization. “AI is taking a human judgment problem and converting it to a machine prediction problem. AI is commoditizing prediction in the same way computers commoditized arithmetic.”

In a similar vein, James highlights that the democratization of AI offers the chance to further unlock new capabilities, enhance customer experiences, and simplify tasks for associates both within and outside the company. In many instances, it empowers them to focus on higher-value activities.

However, he emphasizes the importance of effective management. Democratization necessitates the implementation of inline controls throughout the organization. Without these controls in place, true democratization cannot be achieved.

Responding to Nixon’s question about ensuring discretion among business users while using various AI tools, James emphasizes the importance of establishing clear processes. He explains, "You have to make it clear that there is a process and a mechanism to follow if you want to be able to leverage artificial intelligence."

He acknowledges the initial challenges, stating, "At first, the ground rules are pretty burdensome because a lot of it's manual." However, he suggests a more efficient approach through tooling and frameworks to facilitate the process.

Next, James highlights the significance of understanding different use cases and engaging with the business accordingly. He maintains that the right training and guardrails are a critical part of that rollout.

A key benefit of the AI co-pilots James says is the ability to embed the technology into existing quality control routines for uses like contract review or software quality assurance. “I also think there is an appetite and aptitude for co-pilots because it is something people can understand and see the benefit of using. It's those kinds of interactions where people realize this is a tool that's making them better.”

How to get started with AI?

Most organizations are not as well-versed in AI/ML as Kroger. Data leaders across industries today are feeling the pressure to adopt AI across usage scenarios. When asked about the ideal approach to starting out, James mentions “starting small, identifying a use case and a supportive business owner where you do have the data that can be used to drive it.”

He urges teams that lack in-house resources to work with the vendor community and learn from peers and industry groups. “Get a couple of use cases and build advocacy within the organization across the business leaders where you can demonstrate tangible value. Once you get a critical mass, start thinking about how you build for scale.”

Speaking of the ideal department for the initial use cases, James mentions “marketing” as it manages large amounts of data and tends to be aware of analytics. “The use cases, at least initially, are very helpful and informing, but then there's also a path into automated analytics, activating decision points through capabilities like personalization.”

Upskilling, reskilling, hiring — How to build an AI workforce?

James goes on to discuss the key aspects of building a workforce equipped for utilizing AI tools. He emphasizes the need for the ability to navigate future challenges, stating, "You need talent and capability that's going to help you through the future."

Additionally, education, awareness, and engagement at all levels of the organization are crucial factors in the upskilling and reskilling efforts. James underscores the importance of involving users in the development process of AI initiatives, citing personal experiences of working alongside store associates.

He highlights the significance of collaboration between technical and business teams, emphasizing the cross-pollination of skills and knowledge. According to James, such collaborations often lead to more innovative solutions and better understanding between teams.

Regarding organizational structure and stewardship, he prefers data and analytics to be positioned close to the business. “You will start to find natural ownership paths within the business and a lot of times it's happening before you even declare it.”

James advocates for an approach resembling a manufacturing model, where controls and processes are integrated into the workflow for efficiency and manageability. He underscores the importance of striking a balance between control and enablement in implementing AI and analytics solutions. He also stresses the need for deliberate consideration of tools and capabilities that facilitate business operations.

“I want to build a streamlined, efficient advanced analytics ecosystem where we provide flexibility at the edge, with embedded support and control mechanisms.   It actually has to make their life easier,” says James.

A final takeaway for up-and-coming retail data executives

In closing, James leaves aspiring retail data executives with a crucial insight into the “importance of reducing the distance between data and analytic capabilities and business." He highlights the transformative power that comes from aligning technology teams with the business context and vice versa, leading to a surge in innovative ideas and autonomous evolution of products.

Instead of top-down directives, teams organically innovate and evolve analytic and data products. He advocates for team structures grounded in a product mindset, believing this approach will stimulate organic idea generation and propel projects forward.