When I first worked for a club, we didn’t have a single data scientist on staff yet. Looking back, I didn’t realize the opportunity I had to shape the tech stack and development cycle of both groups from the very beginning. If I could go back and do it again, I would definitely establish some working processes that made it easier for both of us. Ultimately the less you work together with your data analysts, the more work it ends up being for you in the long run.
Benefits of Working Together
The goals of the data and systems groups are very similar – to provide accurate data to the front office and coaching staff as fast as possible. This means the hard part is already out of the way, alignment of priorities between groups. Some departments within organizations compete for resources making such alignment more a test of politics and power. With NFL data and systems groups, the main resource struggle is over each other’s time. Data analysts need data, and it’s the responsibility of the systems group to provide it. We all know this isn’t as simple as it sounds – it involves cleaning, importing, storing, securing, and then making it available to those who will make recommendations for decisions needed to move the team forward. Any hang up in this process leads to delays and frustrations. If these two groups are not working together, they are working against each other.
Approaching the data group as a technology department that has it’s own set of needs allows you to build infrastructure that benefits everyone. There are many analysts today who keep the only copy of a model underneath their desk. Presumed benefits of this are speed and agility – it’s very easy to make a change and redeploy the model. However, if we view the analysts as part of our technology group, this suddenly turns into a huge risk. One stolen laptop later that model is suddenly lost. Putting standard tech processes in place such as version control alleviates a single point of failure, as well establishes infrastructure for both data and systems departments.
A side effect of coming together to create process between teams is better communication and relationships. We’re all used to being last in the communication chain which turns a request that would have been simple into a ‘this has to be done right now’ situation. Opening communication channels with your data group ensures you stay on the same page and ahead of the game. Once you understand the scope of their work, it will illuminate ways you can incorporate their models into your apps. Ultimately this elevates visibility for both stakeholders and not only improves your working relationship with the data team, but also encourages trust from the front office.
One Step at a Time
Sure, I can talk about an ideal world where data scientists and software developers work in harmony. But how do we get there? The first step is to talk to them. Hard work is lauded at an NFL franchise, sometimes even at the expense of improvement which leaves easy gains to made in automation. Ask each of your data scientists what they are most worried about. Understanding a problem is the first step to creating a solution.
After you compile a list of issues, do some research. Version control would absolutely allow parallel work streams on data models and therefore can potentially reclaim time. However, you need to consider the overhead of training the group on the use of source control, and the dev cycle process. I’ve always advocated that the best tech stack is the one that satisfies both the needs of the product AND fits the skillset of your team. Conduct research within your group to see if anyone has experience with a certain technology. It could save you loads of time in the long run with support.
And finally, Rome wasn’t built in a day. Deploy small changes during times when impacts are minimal. No one wants to adopt a new technology mid-season and face a 6am call from a coach who suddenly can’t access their reports. A well thought out deployment plan can be the difference between success and failure of technology adoption.
Can’t We All Just Get Along?
Earlier I mentioned that data and systems groups within clubs are already aligned on goals thereby eliminating politics from the scene. In a perfect world perhaps. There will be some that could potentially see working together with the systems group as a threat. They hold information close to their chest and are afraid of change. The best thing to do in this case is convince them you are on their side and want to make their job easier (and yours as an added benefit). For example, automating a manual data cleaning process gives them more time to create studies and conduct research.
Resources are ironically, another obstacle. You’d think within a franchise that has a salary cap of $250 million dollars you’d be able to request a mere $5k for software licensing that will improve everyone’s lives. Franchises are notoriously frugal for any expenditures outside of football operations and player salary. You can get creative with your budget by utilizing software trials before purchasing and searching for vendors that may eliminate other extraneous costs. Always frame your request in terms of the benefits to be gained. Evolving your tech stack without growing costs is a challenge we all face.
And finally, unpredictability plays a major role. You can make all the plans you want to adopt a new technology in the offseason – until you’re facing staff turnover and need to get a new head coach or GM up to speed very quickly on your existing technology. And let’s not forget, adapting your applications to a whole new set of stakeholder needs. It can feel like all progress halts during these times, but that’s the business. Data scientists are going to ask you where the same data is 14 times a week under different head coaches, so all the work you’ve done to improve workflow is still relevant. It just is deprioritized until you can earn the trust of your new front office.
Takeaways
- Understand the data groups’ needs, research solutions and deploy small changes
- Automation will help strengthen your infrastructure and free up time for both teams allowing innovation to advance
- Talking with your data group helps build communication lines and could become a tool to combat politics as an obstacle
