3 Ways to Drive Data Culture on your Team
How to empower your team and cultivate data culture
One part of our jobs as data scientists is to drive data culture on our teams. This includes making sure that data is part of every decision, that we can quantify the impact we’ve made with new launches, and that we are making decisions analytically, to mention just a few.
As a data scientist, you want your stakeholders to care about the work you're doing not only because data-driven decisions are good for the business, but also because that’s how you are driving value. You want data to be included in the planning process, not to be an afterthought. Below are 3 steps to drive data culture on your team that have a track record of success.
Not everyone has prior experience working with data scientists. We have had people shyly ask what it is that we do. We have found it a worthwhile exercise to deliver a presentation to our team talking about the work we do and how we can improve the reliability and value of the products being developed. We had also created a document outlining how to work with us on a daily basis. Being open about your prioritization framework and sharing your quarterly OKRs with stakeholders is also helpful. Presenting cross functional demos is another way to demonstrate value and share knowledge as well.
For example, sharing who is responsible for what during an A/B testing rollout and analysis process ensures that no task falls through the cracks due to the lack of ownership. Aligning on what we will be evaluating ahead of time can prevent PMs asking questions about metrics that we aren’t currently tracking.
We ask stakeholders to ask business questions, instead of data questions, and share the context. Sometimes, they may ask for the wrong data or their request may be computationally expensive and time consuming. As data scientists, we can help them come up with proxies. We also want to make sure that the data or analysis provided will make a business impact, not just answer a curiosity.
Show stakeholders the value of data by proactively uncovering new opportunities and identifying issues
It’s important to have an alerting system in place for when your metrics move drastically, whether that’s in a positive or negative direction. This could be automated and sent via email or Slack to make your life easier. If that’s not possible, you should look at your dashboards regularly to proactively identify issues.
As an example, if the volume of users onboarding drastically changes, you could see whether that’s an issue with your ETL, the underlying tracking, or whether there’s been a change introduced by a product team that led to the change. On the Trusted Identity team, we’ve noticed changes in our onboarding funnel and, after investigation, we’ve realized this was due to the changes introduced by another team. This resulted in a discussion about better collaboration going forward and proactive communication across both teams via a dedicated stakeholder Slack channel.
You want to lead by sharing what’s going on with your stakeholders and that you are investigating the changes. Once you have the conclusion of your investigation, you should share that with your team as soon as possible and make recommendations for next steps.
Standardize and ask for a metrics section in the PRDs (Product Requirements Documents) and Engineering Design Documents
A metrics section in the PRDs will ensure that there is a way of measuring impact of new launches. You should sign off on it to make sure that the metrics truly reflect the measurement of what’s being changed. This will also allow you to confirm that creating the metrics is possible and that you have a way of measuring them.
A similar section in the engineering design docs will allow you to design logging and ensure that engineering implements the necessary tracking required for measuring the success of the launch.
Having these metrics sections has made our product and data science partnerships even stronger. We have a discussion about what the goal of each launch is and how to measure it. This encourages everyone to think deeply about what we care about and go beyond our main metrics, such as onboarding funnel. In the past, we have uncovered that while the funnel wasn’t affected by the new experience, it took users a longer time to complete it. This wouldn’t be possible without us thinking through the metrics for the launch.
For a data scientist, driving data culture on your team is one of the most valuable things you can do. Encouraging data driven decision making will lead to better business outcomes and will improve working relationships with your stakeholders.