5 years, I left Brex in August. Time actually flies. I joined Brex when it was three years previous, and I can nonetheless vividly keep in mind my first day. On the identical time, these 5 years felt for much longer. That is all as a result of in depth information I’ve gained alongside the way in which, from becoming a member of a younger knowledge workforce and studying the complete knowledge cycle, to collaborating with Operations, GTM, and Product, and ultimately transitioning into management. This accelerated progress is without doubt one of the largest advantages of becoming a member of a startup as a knowledge scientist.
“Shall I be part of a startup or a extra established firm?” I hear this query from my community on a regular basis. On this article, I’ll share the information scientist’s life at a startup based mostly alone expertise. Hopefully, this may show you how to navigate by means of your individual profession path.
A fast be aware: “startup” is a broad time period that covers many kinds of firms. For context, Brex is a US-based fintech startup that gives company playing cards, banking companies, and expense administration software program for enterprise clients. I joined Brex after three years of full-time expertise in product knowledge science. On the time, Brex was a three-year-old Sequence C startup with 500+ workers. Subsequently, my expertise might differ from what you’d expertise at an earlier-stage firm.
I. Area-specific vs. Purpose-oriented
Startups transfer quick — they should regulate methods based mostly available on the market suggestions and iterate continually to make sure product-market match. Information Scientists, as an costly useful resource, have to shift their focus space accordingly to create the most important enterprise worth.
Taking my very own expertise for example:
- After I first joined Brex, the precedence was to onboard as many shoppers as we may whereas controlling for fraud threat. Subsequently, I labored carefully with the Onboarding workforce to optimize the account utility circulation and enhance useful resource allocation for the Operations workforce.
- As our buyer base grew, the subsequent precedence was to scale our buyer help (CX) perform to offer the best-in-class help. I partnered with the CX workforce to investigate buyer ache factors and cut back product frictions. (Learn my article for extra particulars on DS in CX.)
- Later, I collaborated with the Implementation workforce to speed up ramp-up for brand spanking new clients and with the product workforce to establish churn drivers and enhance retention.
As you possibly can see, startups are very goal-oriented. At massive companies, you may specialise in one product function for years; At a startup, the corporate’s wants can fully change inside 1 / 4, and knowledge scientists change their domains extra continuously consequently.
The upside is that this permits the information scientists to discover and collaborate with extra features and have a extra holistic view of the enterprise. I labored with practically each workforce at Brex previously 5 years — watching how GTM (go-to-market) discovered prospects and closed offers, seeing how Product and Buyer Assist labored collectively to handle buyer ache factors, and studying how the Buyer Success workforce diminished churn and drove upsell. It is a far more practical technique to perceive enterprise operations than studying books or taking programs at a enterprise faculty. It additionally ensures the information scientists have clear and high-impact enterprise questions to unravel.
However what’s the problem? Information scientists have to comply with the corporate’s technique and adapt to modifications continually. This implies shifting priorities accordingly, and generally having to construct issues the quickest means, as a substitute of essentially the most scalable or dependable means. Tech debt piles up shortly with tons of one-off dashboards and a number of variations of the identical metric, complicated stakeholders and the information workforce itself. Establishing the perfect practices takes time, so residing with the chaos generally is unavoidable — although having the ability to arrange the requirements from 0 to 1 (or seeing others doing so) can also be a rewarding a part of working at a startup, particularly for aspiring knowledge science leaders.
II. Information Analyst vs. Information Engineer vs. Information Scientist? All of the Above.
Job titles in knowledge today are complicated — Some Information Scientists deal with experimentation, whereas some are deep in machine studying; Some Information Analysts merely construct dashboards, however some primarily do product analytics. However relating to startups, titles matter much less — you’ll seemingly do every little thing.
After I joined Brex, everybody had the identical title, “Information Scientist”, however all of us needed to put on a number of hats. We cut up the information workforce into DS, DA, and DE features solely since early final yr. As I discussed above, startups are goal-oriented. Subsequently, when there are restricted folks and no clear workforce construction, you will have to do every little thing to realize the objective.
- Information Engineering: I realized data modeling, Airflow, ETL processes, SQL optimization, and plenty of different knowledge engineering expertise previously 5 years. I keep in mind considered one of my first tasks at Brex was emigrate our knowledge pipelines from dbt to an inside software. Sure, as a knowledge scientist at a startup, there’s a excessive probability you additionally have to construct, personal, and keep your individual knowledge pipelines.
- Analytics: Becoming a member of a startup means there are numerous enterprise areas with no or minimal knowledge help. Subsequently, to assist the workforce higher perceive their efficiency and acquire their belief, the very first step (after you’ve constructed the information pipelines) is to outline the success metrics and construct the dashboards. As soon as the metrics are carefully monitored, questions like why they moved or easy methods to transfer them come subsequent naturally. These are all frequent analytics duties.
- Information Science: There are additionally loads of superior knowledge science use circumstances at startups. Machine studying fashions are useful to detect fraud, predict churn, estimate LTV (lifetime worth), and so on. Causal inference can also be vital to judge the impression of a advertising and marketing marketing campaign or a product launch.
Is it a very good or dangerous factor to put on a number of hats?
If you’re not but positive about which sort of knowledge monitor you’re most excited about, becoming a member of a startup will show you how to to get some taste of every little thing and determine which path to go subsequent. Or if you want to develop into a Head of Information and even construct your individual firm sometime, this full-cycle publicity is tremendous invaluable.
In the meantime, the draw back can also be very clear — you’ll spend time on issues you aren’t excited about, or not that related to your long-term profession objective. The truth that you personal the entire knowledge lifecycle may confuse stakeholders, as they often solely care a few sure kind of output, for instance, dashboards or fashions, however don’t understand it’s essential to spend one other 50% of time on knowledge pipelining.
III. Greater visibility
Information scientists at startups by no means lack visibility. From day one at Brex, I started working straight with the C-suite. Management will typically come to you with pressing and essential enterprise questions, hoping you are able to do some knowledge magic to uncover insights and drive enterprise progress. This isn’t one thing you often get to expertise at a longtime firm, particularly as a junior knowledge scientist. It’s a high-pressure however extremely rewarding surroundings.
For instance, through the Silicon Valley Bank Crisis in March 2023, many startups have been impacted, going through the chance of shedding their operational funds. I labored very carefully with the management workforce to assist startups survive this difficult time. I created a real-time tracker on new buyer purposes, analyzed utility evaluation velocity to estimate extra workforce wants, and collaborated with different DS to automate and velocity up onboarding checks. It was an intense weekend, working cross-functionally in a warfare room (nearly on Zoom) from 8 am to midnight. Nevertheless, that is considered one of my finest reminiscences at Brex, exhibiting the true buyer obsession from our management, and the way knowledge scientists can contribute straight and drive big enterprise impression.
IV. Publicity to new instruments
Younger startups are additionally courageous sufficient to strive a brand new tech stack. Subsequently, they’re typically the early adopters of latest instruments, whereas bigger firms may take months (and layers of approval) earlier than even piloting, to not point out the numerous migration prices.
For instance, throughout my time at Brex,
- We began exploring LLM use circumstances in knowledge science two years in the past, and have entry to all main LLM APIs (OpenAI, Claude, and Gemini) internally, open to everybody. Yearly, Brex hosts an inside hackathon and encourages workers to innovate. Two years in the past, I collaborated with engineers to construct an AI-powered customer feedback platform to mechanically categorize, summarize, and analyze numerous unstructured suggestions knowledge. Final yr, I constructed a RAG-based chatbot to assist stakeholders retrieve buyer suggestions associated to a selected product function simply. This yr, I labored with different knowledge scientists to discover text-to-SQL capabilities with Claude Code and Snowflake CLI. I completely loved these alternatives to use cutting-edge methods to knowledge science workstreams.
- We continuously piloted new knowledge options. For instance, we have been an early buyer of Hex for collaborative and easy-to-share knowledge notebooks. We use Statsig for experimentation and occasion monitoring. We additionally tried numerous AI-powered enterprise intelligence software program for higher self-service analytics.
Working at a startup helped me keep knowledgeable of latest applied sciences and undertake them into my each day workflows. That not solely made the work extra thrilling but in addition stored me aggressive because the trade advanced.
On the flip aspect, being an early adopter can imply disrupting present workflows and rebuilding infrastructure. It additionally means a much less steady improvement expertise.
Conclusion
So, must you be part of a startup or a extra established firm?
Startups supply velocity, selection, visibility, and cutting-edge publicity. However in addition they carry chaos, shifting priorities, and the necessity to put on hats you might not get pleasure from. When you thrive in ambiguity and need to speed up your studying curve, a startup could be an unbelievable place to develop as a knowledge scientist.
For me, 5 years at Brex have taught me an unbelievable quantity of data about enterprise and knowledge. I’ll perpetually be pleased about the teachings, the folks, and the prospect to see what knowledge science can seem like at a fast-growing startup.
