How I got here
I had a marketing background - product marketing, event marketing, some affiliate stuff. I was not good on the creative side. My boss redlined 99% of my writing. So I moved into Google Analytics, then operations analytics, then someone offered to teach me SQL and I said yes because how often do you get paid to learn?
My advice to anyone trying to get into data: say yes to every task. Everyone needs a report done somewhere. Just do it where you're already working and figure out how to grow into data from there.
Building from zero
When I joined Clearbit at 40 people, there was no consolidated data. No single table of all users. No attribution. No funnel metrics. The first thing I did was call Tristan and say we need help. We brought in a consultant and started building the models - consolidating users, building out funnel reporting, web analytics. All of it had to be built.
I use 80% as my threshold. There's no scientific reason behind it, but once I could answer 80% of the questions people were asking on a consistent basis, I moved to the next stage.
The report graveyard
I fumbled for about a year building complex dashboards. The amount of reports sitting in what I call the report graveyard is embarrassing. Hours and hours building dynamic parameters so someone could answer any question they could ever possibly think of. Nobody used them.
The lesson: go simple. Answer the easiest question everyone has. Once everyone can answer that, move to the next one.
Self-serve means meeting people where they are
We spent money on Mode and I tried to make it work for everything. Our sales leaders just weren't going into Mode. So we put the data where the GTM teams already lived - Salesforce. We reduced one barrier to getting the data by not asking people to learn a new tool.
The first time I felt truly successful was when the demand team could answer questions about which campaigns drove the most pipeline without ever asking me. They had attribution, they had influence data, and they never needed me to build them a report. That's self-serve.
80% for reporting. Not for operations.
As we started pushing data into more places - Salesforce, Customer.io, marketing automation - I went from being at the end of the process to being in the middle of it. More exciting, more impact on the business, but higher risk.
80% is fine for reporting. Nobody's going to accept 80% on a financial report, obviously, but for most internal dashboards it gives you direction. 80% is not fine when you're sending emails. 20% wrong means 20% of your customers get bad data in their inbox. That gets you on Twitter for the wrong reasons.
More places you push data, more chances it breaks. You change a dbt model and forget you're syncing it to a Google Sheet somewhere - now that report is out of date and nobody notices until someone goes to use it.
What's next
At the time of recording, we had BigQuery, Snowflake, and Redshift all running in parallel. The goal wasn't necessarily one warehouse. It was one set of rules, naming conventions, and processes no matter where the data lived. Apply dbt so everything follows a similar path regardless of how it was collected or where it's stored.
My hope for the industry: data teams become part of the action early - in the brainstorming, the campaign creation, the product decisions. Not coming in three-quarters of the way through to report on what already happened.