Ecommerce brands rarely struggle with a lack of data, but what to do with it all as they grow and complexity creeps into their workflows.
Cary Lawrence, CEO of Decile, says the issue becomes clear once brands move beyond early-stage reporting. Native dashboards inside ecommerce and marketing platforms work at first, but each tool only tells part of the story.
“As a brand starts to grow and their ecommerce stack evolves, the native dashboards don’t cut it anymore. Each tool provides a set of reports from their perspective, but there’s a gap between them,” Lawrence explains. “There is an inherent need to translate and stitch the data. Teams spend hours downloading and stitching together reports from various platforms, but even then the results often don’t show the why.”
Marketing teams often try to close those gaps manually, exporting data and assembling reports that still fail to explain what changed or which actions mattered. Context disappears when metrics live in isolation, leaving teams with outcomes but little insight into customer behavior.
Why Decile built Luma differently
The rise of AI promised to simplify analytics, and many ecommerce teams have started experimenting with general-purpose AI tools to summarize reports or query exported data. Results often fall short and lack the depth teams need to act with confidence.
Lawrence says the gap comes down to how ecommerce data actually works. “The questions that sophisticated brands are asking are much more nuanced,” she explains. “While the use cases are simple, the technical elements are complex.”
Generic AI models struggle with that complexity because ecommerce data carries meaning that changes by brand, channel, and vertical. Order behavior, returns, discounting, subscriptions, and lifecycle metrics do not follow a single set of rules. Without domain-specific grounding, AI has difficulty producing reliable answers.
Decile took a different approach when building Luma AI. Instead of adapting a general AI solution, the company built Luma on a proprietary ecommerce data model shaped by years of direct experience working with brands.
“We didn’t start totally from scratch,” Lawrence says. “We provided our proprietary ecommerce expertise, informed by more than five years of work with ecommerce brands, as additional context for the enterprise-grade LLMs we’re using for Luma. Since Decile already had data warehoused in Snowflake, we were able to leverage Snowflake’s Cortex stack to enable the technical components.”
Together, the data foundation and ecommerce-specific context allowed Decile to build an analytics agent designed around how ecommerce businesses actually operate. Rather than treating metrics as generic data points, Luma evaluates them in context, translating questions into analysis grounded in each brand’s structure and customer behavior.
Why transparency changed how teams trusted AI
Trust remains one of the biggest barriers to adopting AI-driven analytics. Teams hesitate to act on insights they cannot validate and confidence erodes quickly when recommendations arrive without explanation.
Luma addressed this directly by making its reasoning visible. Users can see how insights were generated and which data informed them.
“Including the ‘thinking’ portion of Luma ended up having a huge impact on our client’s trust and the legitimization of Luma AI,” Lawrence says. “We had no idea how powerful this functionality was going to be until we kept hearing this theme come up over and over again from our initial private preview clients. Understanding the ‘how’ for the insights and recommendations coming from Luma was super powerful, regardless of whether the user was a marketer or an analyst.”
Seeing the reasoning behind each insight removed hesitation across teams. Marketers stopped second-guessing results, and analysts gained clarity into how conclusions were formed. Transparency played a key role in building trust around AI-driven analysis.
How AI analysts will shape future ecommerce teams
Looking ahead, Lawrence expects AI analysts to change how ecommerce organizations operate. Rather than limiting analysis to a business intelligence team, future organizations will treat data fluency as a shared capability.
Broad access to trusted insights allows teams to move faster without adding technical headcount. Decision-making shifts closer to the work itself, removing the friction that once came from exporting data, reconciling reports, or waiting on specialized resources. First-party customer data becomes the strategic advantage teams rely on day to day.
For Lawrence, the goal remains straightforward. Analytics should support action, not slow it down. When insights become accessible and trusted, data finally works the way ecommerce teams have long expected it to.
