I run data and technology at Hydration Room and co-run Hexmark with Colby Schrauth.
What a data leader is for
Building is getting cheaper every month. An analytics engineer working with Claude can build most of what a business asks for, so the scarce skill is judgment, turning a business problem into a decision and pushing back on the request when it needs it. What people ask for literally and what they are actually angling for are often two completely different things. A lot of data professionals can’t hang with a CEO. They’re builders, not business strategists, and the ones who stay that way will struggle in the years to come. I wrote more about this in Moving up the value chain.
Data work is business work
Writing code and getting data to “answer” a question is the immaterial part of the work. The first version of almost any model comes together quickly. What takes months is everything around it: learning how the business actually runs day to day, understanding why the numbers look the way they do, and working with the people closest to the work on what should change. More often than not, the change that matters most is an operational one. Done well, data work is operational change, and operational change depends on clear communication, on the trust of the people who run the business, and on being close enough to them to find the subtle nuances. That is the part of the job that falls to a leader.
AI readiness is data readiness
AI is probabilistic. It drifts and it sounds confident, and if you let it, it will create fallbacks that hide its own errors. Working with it feels like running a data team whose B squad is very fast but still needs a senior to check the work. For anything consequential, I’d rather have a deterministic answer from the warehouse than a model’s best guess, because the warehouse tells you the same thing every time. Storage is cheap and tokens aren’t. Getting a business ready for AI mostly means giving the model metric definitions it can follow through the business logic, so it isn’t guessing from dashboards.
Centralize the definitions
Every tool that makes analysis easier pushes it out to the edges of the company. Excel did it in the late ’90s, and self-serve BI did it again. I spent years on that second wave, rolling out Tableau and Power BI and watching dashboards proliferate like rabbits. AI is the biggest version yet. Each wave has ended with the business pulling its definitions back to the center, because more analysis only helps when everyone agrees on what a number means. I want the questions spread as widely as possible and the definitions kept in one governed place. I wrote about the AI version of this in The spreadsheet era, again.
Buy most of it
No one suggests you bake your own bread every day or build your own furniture whenever you need a new couch, and software should be the same, with caveats. AI makes building your own tools look cheap, and the companies going all in on it are going to find there was a reason they paid people to do this to begin with. I’d rather buy what a vendor maintains well, push vendors to give the business access to its own data, and build a small tool only where nobody sells one. Whatever we do build lives in the business’s own accounts.
Security that enables the business
The goal of security should be to safely get the business to its stated objectives. Too often it puts the onus on the business to prove its plans are secure, and people work around security that feels onerous, which creates exactly the risk it was meant to prevent. With development this fast, the businesses best positioned are the ones whose security teams say, if you want to do this, we’ll figure out how to make it safe, and then hold strict defaults so the safe path is the easy one. More in Security’s culture problem.
2025 to now
Hydration Room, PRIORA and Hexmark
In October 2026 I joined Hydration Room as Director of Data & Technology, after working with them as a consultant through PRIORA. I own the data warehouse and the reporting our clinic and regional leaders use day to day. AI agents now do much of the development work, and I review what they build before it ships.
Colby Schrauth and I run PRIORA, a fractional data practice, and are building Hexmark, a next generation site selection tool for small and medium businesses. Hexmark started as a map we built to help a client decide where to open their next location. It worked better than the site reports they had been paying for, so we turned it into a product.
2019 to 2025
Daxko and Sequoia
When COVID hit, gyms were closed and Daxko, which makes software for gyms, clubs and YMCAs, was scrambling to adjust. I led the development of an analytics offering that doubled as a value-added service for our customers, and it grew past $1M in ARR. I also led the move of our customer reporting from Birst to Power BI. Daxko is where my career had its most important inflection point.
At Sequoia I put what I had learned at Daxko into practice, reporting to Colby and leading data visualization and analytics on Fivetran, Snowflake, dbt and Power BI for more than 1,200 people. The bigger change was in the job itself. Technical proficiency is the obsession of perhaps the majority of ICs in the tech world, but efficient code and beautiful data models do very little to move the bottom line on their own, and data teams that stop there end up viewed as (expensive) cost centers producing dashboards that proliferate like rabbits. At Sequoia I moved up the value chain, from a worker bee taking requests to an advisor on the economics of the business. I learned its unit economics, spent time with department heads in their world, and worked closely with finance on the P&L, and I judged the data work by whether it moved the bottom line.
2016 to 2019
Dish Network and myStrength
At Dish Network I built data marts and self-serve Tableau reporting. At myStrength, a mental health app, I ran the database schemas, ETL and reporting server, automated the monthly and quarterly reports, and co-authored a peer-reviewed paper on outcomes.
2,165 commits over the past six months across 18 repositories, from my personal account, lukesnow2, my PRIORA account, luke-priora, and my Hydration Room account, lsnow-hydrationroom. Mostly Python, TypeScript and SQL. Most of that is client work in private repositories. Below are a few of my own projects.Counted October 18, 2026.
Hexmark
Site selection for small and medium multi-location businesses. Describe your customer and it ranks the neighborhoods worth a visit across a metro, then scores any address on the people nearby, the competition and the overlap with your own stores.
Every season of our high school fantasy football league since 2005, pulled from Yahoo into a Postgres warehouse and served as a web app. It loads the new week on its own every Wednesday during the season.
A read-only MCP server that lets an AI assistant search Colorado library catalogs and cite real records instead of making up titles. I built it to help find leads on terroir-style cooking along the Front Range, on the West Slope and in Colorado's mountains.
Not public yet
Santa Barbara
2009 to 2016
AppFolio
I had planned on law school after UC Santa Barbara, but the legal market was quite poor in the years after the Great Recession, so I took a job at AppFolio in Goleta to bide my time. It quickly sparked in me the joy of using data and statistics to fix what I saw as deep procedural flaws in how my team operated. I interviewed with the team’s VP to move into a business analyst role and demoed a few analyses I had run on team performance. I was told in no uncertain terms that I wasn’t qualified, not because of the analyses but because I had no degree in statistics and no experience in business analytics. Within three months I had an offer from Dish Network in Denver.