Arcee is a fast-growing AI company building high-performance generative AI models that organizations can fully own, customize, and deploy within their own infrastructure.
In a rapidly changing AI space, Arcee hit the ground sprinting, with constant iterations, product launches, and investor pitches. Like many startups, early tooling was optimized for engineering speed, not for product visibility, go-to-market insights, or leadership decision-making.
Davis Stone, Arcee’s Head of Growth, describes their initial customer data stack as “accidental” at best.
In the beginning, Arcee was using whatever tooling was already in place, which in their case was PostHog, a tool that worked well for the engineering team for various product insights. As marketing and growth became a priority, they tried to use that same platform for go-to-market analytics, extending it past its limits and repurposing a product-focused tool to answer questions it wasn’t designed to handle.
For a while, they pushed through.
“...For a while, the sentiment was ‘Let's just figure it out. We're engineers, it's a startup, we build stuff’. But the product itself is what we needed to build. And now all of a sudden we're pulling people away from that to try to report on the product.” – Davis Stone, Head of Growth
As the company scaled, cracks began to show. Product engineers became the de facto data team. Every one-off insight was on borrowed time.
.png)
This creates two problems at once:
That reality became clear at Nvidia GTC. Arcee had just launched a new product and was actively promoting it at the conference, but had no visibility into how it was performing.
Davis remembers pulling a field engineer aside, hunting for WiFi, and asking him to help write SQL queries on the fly. The engineer had never used PostHog before. They were using Claude to generate queries and trying to translate them into something that would run, just to get basic launch insights.
The team was flying blind. Without a feedback loop, Arcee couldn’t tell what was working, who was engaging, or why momentum stalled. The lack of insights directly impacted product outcomes.
Trying to gain insights felt like “little paper cuts” that multiply exponentially, and the system was strained. This is one of the earliest warning signs that a company has outgrown its data approach.
“I could see the product… die on the vine because we couldn’t understand what was happening.”
At the same time, Arcee was pivoting from a sales-led to a product-led growth strategy. Sales-led or founder-led approach involves high touch, white-glove action with relationship-driven, one-on-one interactions to close high-value deals. Early sales calls are “feedback calls” for grasping what customers need in a product and understanding the sticking points.
As Arcee was getting ready to scale up to a PLG growth motion, they knew they’d have less visibility into their customer base, what products they were using, and how. That’s when data became critical for staying informed on customer needs.
“We needed a full-service data team, immediately.”
Arcee needed execution, context, and speed. A hands-on team with end-to-end solutions. They needed someone who could work within strategic framing and move quickly at the same time.
Data Culture joined Arcee during a period of rapid transformation and discovery.
Working closely with the Growth team, Data Culture was able to quickly integrate as part of the team, parallel to Arcee’s fast-paced product iterations.
"Definitely something that stood out about working with Data Culture is moving at a startup pace. That was seamless. That was how you all work, while also bringing a level of stability and confidence and calm to an otherwise unstable and un-calm space."
Data Culture made a point to know the context, knowing how the product worked, who Arcee’s customers were, and how to reach them. They became an extension of Arcee’s own team, with an active Slack channel that the entire organization used to pose questions and give suggestions.
Rather than over-engineering or over-planning, Data Culture helped Arcee define a minimum viable data foundation.
This foundation could evolve as strategy changes, without creating technical debt they’d regret later. They helped build something flexible, without cutting corners.
.png)
And when Arcee hired a head of product, it was easy for her to instantly jump in and be effective, thanks to the data foundation Data Culture had in place.
As Arcee evolved from a PLG motion to an open-source ecosystem player, its data strategy matured alongside it.
Today, data drives product iteration, GTM prioritization, partner strategy, and executive storytelling, without pulling engineering away from building the product.
Arcee now uses Omni’s AI analytics platform and Snowflake, tools that took Arcee to the next level. This allowed for deep analysis, insightful dashboards, and data insights that tap into every aspect of their product, from how their product is being used to who is using it, to informing their distribution strategy.
Just as important as what was built is how it was built. The platform was built for what comes next. It was built to evolve, knowing the nature of the team and environment will shift and change quickly
As a startup, they were always experimenting with different strategies and needed to move fast in AI spaces. Data Culture built an infrastructure that was as agile and flexible as Arcee is.
The outcome isn’t just better reporting, it's:
As Arcee has doubled down on expanding its distribution channels through partner networks, data sharing has become much more crucial. Snowflake’s data share capabilities has enabled them to get the data they need from partners easily without a heavy engineering setup. It’s made data sharing much easier with all the patterns Arcee is working with in the AI ecosystem.
And using tools like Omni Analytics, everyone at Arcee now has access to the insights they need, with customizable views, flexible querying, and drill-downs that can inform decisions and steer iteration planning across teams.
They had context, not just numbers, which made it simpler to hold partners accountable and celebrate wins with them. Easily answering why a toolkit, specific model, or distribution channel fits together for their clients.
For high-growth startups, the Arcee story serves as a blueprint for moving from an "accidental" data structure to strategic intelligence platform.
Avoid the “Part-Time Data Engineer" Trap: When engineers are on borrowed time finding data insights, they are pulled away from their actual work, creating de-facto data analysts who need to split their time. Investing in a data team early protects a valuable resource: engineering hours.
Play Out the Hypothetical: Davis suggests asking, "What does this look like in six months? In a year?" If current tooling requires a custom build for every inquiry, start-ups begin building technical debt, not a data stack that can grow with data needs.
Context is Everything: A data partner should do more than report numbers; they should understand your go-to-market motion, your product experience, and your customers. As Davis noted, the biggest wins happened when the data team understood why. This way, asking to expand Hugging Face reports to include OpenRouter is met with understanding, instead of an entire meeting to explain the ask.
Focus on Ecosystem Impact to Shape Narratives: In the AI frontier, traditional metrics like DAU don't tell the whole story. Startups need the infrastructure to track "macro trends" and ecosystem engagement to stay credible with investors and partners.
