So You Wanna Keep Your Job in Data Science… Part 2
Welcome to another edition of “In the Minds of Our Analysts.”
At System2, we foster a culture of encouraging our team to express their thoughts, investigate, pen down, and share their perspectives on various topics. This series provides a space for our analysts to showcase their insights.
All opinions expressed by System2 employees and their guests are solely their own and do not reflect the opinions of System2. This post is for informational purposes only and should not be relied upon as a basis for investment decisions. Clients of System2 may maintain positions in the securities discussed in this post.
Today’s post was written by Kevin Nutter with input from Derek Liu and Young Se Choi.
Part 1 of this series is available here.
Building an indispensable data science department requires much more than great data science. Contrary to an in-house data science team, which only needs to build one successful data science department, at System2, we build a new department for every client. We’ve landed on five things that must happen to manage and develop a data science department effectively:
Prioritize - actionable analyses above informative ones.
Track - the time the data science team spends on different analyses.
Score - ask the client’s team for a score on each analysis.
Present - communicate the scores and efficiency to the senior stakeholders.
Iterate - optimize the data science team’s time on the activities with the most impact.
I’m diving into each of these steps in more detail through 2024. For today, we’ll focus on tracking our work.
Why? To ensure our time spent optimizes the fine balance of keeping a steady flow of low complexity, high probability of success, and informative projects, combined with a few high complexity, novel, and differentiated analyses.
Friday afternoons are often a time to wind down as the work week draws to a close. For some, perhaps the time comes with the promise of happy hour on the horizon. But amidst the allure of discounted drinks and appetizers, there lies a hidden opportunity for data science teams to collect data to generate insights: submitting timecards.
As a case study to highlight one of the many benefits of tracking time in the context of our prioritization matrix, we examined our time spent on one client over the last six weeks.
Early warning system for problematic projects: dropped spend on highly complex but only informative projects from ~20% to 0%
In February, we noticed a concerning trend of material time spent on informative and high-complexity projects (green line). We reviewed this, highlighted the project during our weekly prioritization call on the week of March 3rd, and made a transparent and collaborative decision with the investment team on whether to continue. In this case, we stopped all work on it as soon as possible, and by March 24th, we spent no additional time on the project.
The derivative takeaway is that problematic projects don’t just waste time; they have a material opportunity cost. The time was then shifted to actionable projects, moving from 25%-35% in Feb to 60-70% in March.
Spend time where it’s valuable: System2 spent ~60-70% on actionable projects in March
After we stopped working on our problematic project, we saw our ratio of time spent starting to return to our desired breakdown. There are always weekly fluctuations, but when evaluating each client, we’ve found client relationships are harmonious when we’re averaging 60%-70% of our time on actionable projects and 30%-40% of the time on informative projects.
As a further breakdown, in the figure below, we also like to see the blue line above the yellow line. We want to spend the most on Actionable-High Complexity projects as they lead to the most differentiated insights.
We’re data scientists, so by nature, we like data. What better data to analyze than our own? That’s exactly what we do here at System2. We’ve tracked more than 100,000 hours in the last few years. For System2 data scientists, submitting a time card every Friday is not just a bureaucratic obligation; it’s a crucial step in analyzing the efficiency and success of our data science teams.