What Full-Stack Analyst actually means here

Phil leads the data team. He spent four years in marketing analytics at DraftKings, earned his MBA at Berkeley Haas, and worked in pricing analytics at Lyft before joining Cardless to build out the data function. Outside of work, you can find him cycling the hills of San Francisco or walking his dog.

At most companies, answering one data question is a relay race: analytics engineering owns the pipeline, the BI team owns the dashboards, someone else owns the analysis. A question comes in, gets passed down the line, and an answer falls out the other end a week later. Everyone stays in their lane.

We designed our data team around full-stack analysts, on purpose. One person takes a question from the raw event in the warehouse all the way to the recommendation someone acts on. The person who modeled the data is the one who tells you what to do about it.

In practice, "full-stack" spans extraction and loading, the warehouse, transformation and modeling, the semantic layer, delivery, and increasingly, AI tooling on top of all of it. At a bigger company, those might belong to three or four different teams. Here, one person follows the thread end to end. Not because we can't hire specialists, but because the work is better when one person stays close to the whole problem.

An analyst joined us from an analytics-engineering background: deep on pipelines and modeling, less experience on the reporting side. Two weeks in, the team needed a dashboard for leadership, but the underlying data wasn't in any condition to support it. There was nobody to hand either half to, so she fixed the data and built the dashboard on top of it, working directly with stakeholders the whole way through. She couldn't have built the dashboard without fixing the data first. Those were never actually separate steps.

AI is making that kind of range more practical. You don't need to arrive as an expert in every layer of the stack. You do need to be able to get productive quickly in a layer you've never touched before.

When we migrated our data warehouse from Redshift to Snowflake, we were quoted tens of thousands of dollars and months of work. Two analysts did it in a week, while keeping up with everything else on their plates, using AI to work through unfamiliar syntax whenever they hit a wall.

It only works because of who we hire.

The people who thrive here have a pattern: somewhere in their past, they jumped into something they'd never done, figured it out, and shipped it. Everyone says they're willing to learn. What I look for is evidence that someone already did.

Missing skills won't make you unsuccessful here, but drawing hard lines around what you're willing to touch will. "I only work on the backend" and "I don't build dashboards" are reasonable positions at a bigger company. Here, the work on the other side of that line doesn't get reassigned to someone else. That's a tradeoff we've made deliberately.

Another analyst joined with zero payments experience. Two months in, he was leading a deep dive on transaction declines and talking directly with a card network. The underlying data is esoteric, and the pipeline to get at it didn't exist yet, so he built it. Then he turned the output into something our partners could actually use. He walked in without the domain knowledge or the data foundation, and built both on the way to the answer.

These are great analysts, but what made them effective here was range: the problems shift weekly, the infrastructure is always a work in progress, and nobody is coming to do the part you don't want to do. That's a real tradeoff, and it's not for everyone.

If that sounds like the kind of work you want to do, we're hiring.