Successful Data Scientists are Empathetic Guides
Data Scientists might reconceptualize their role as an empathetic Data Guide to be more impactful.
Good Data Science is most impactful if the output is interpretable and relatable to the people making business decisions. Data Scientists consider their work to be largely quantitative, however we need to consider why data artifacts are created to successfully shape decisions. This post explores how Data Scientists might reconceptualize their role as an empathetic Data Guide to be more impactful.
What is the Role of Data Science in Product Development?
Data Science means different things to different practitioners. Here I will focus on Data Science in the context of product development.
Thanks for reading people + products! Subscribe for free to receive new posts and support my work.
Data science methods are great tools for getting to know our business and making progress to our goals. However, they are meaningless in isolation from Marketing, User Research, User Experience and Product team context and understanding. Data Scientists benefit from reconceptualizing their role in decision making from a quantitative pillar of knowledge to an organization that listens deeply and guides everyone to the right path.
Effective Data Scientists are Guides
I like to think of myself as a Data Guide that has been added to a team of explorers trying to find their way to the top of a mountain to hit a big ambitious goal. My cross-functional team knows where they want to go. This might be a high-level strategy like “Growing the Business” and as a guide I can put a meaningful metric on this goal such as increasing Daily Active Users or Depth of Engagement. Once a metric is established that represents their goals, the trailhead becomes clearer and this metric can guide their progress throughout the year.
The work is not done when the team identifies the trailhead, Data Scientists must continue to guide good product decisions as they explore the twists and turns that happen during development. They might hypothesise that changing the order of onboarding cards will improve new user engagement, however they don’t know how to evaluate the potential of taking this path vs others. Luckily, they have brought their experienced mountain guide with them and Data Scientists can assess the terrain to avoid common pitfalls. Even on a mountain we haven’t climbed before Data Scientists are familiar with the types of terrain we will encounter along the way. We use data tools such as key results, exploratory data analysis, opportunity sizing and A/B testing to find the best routes.
When I approach my Data Science job from this perspective I am no longer primarily focused on how my deep knowledge of data is represented to my team, instead I am leaning in to cross-functional communication to be successful. The work of Data Science is unclear without cultivating the right context to respond to the right question. I don’t lead way out in front by establishing the goals and strategy. The product team knows where they want to go, they just need help figuring out how to get there. I also don’t follow way behind rotely completing each specific request I get for insights. Instead, I walk with them listening to their objectives in context and guiding them with the most relevant data to get there.
Data Maps and other Artifacts
The primary output of Product Data Science is a meaningful artifact that will guide cross-functional stakeholders to make good decisions about the business. However, when I speak with Data Scientists they oftentimes are focused on how their artifacts will appear to other Data Scientists. Is the finding reproducible? Is the code in GitHub? Can we understand what parameters were tested and what the results were? Can my notebook make a rigorous report for our repository?
When Data Scientists focus on these questions, the artifacts we are creating are not considering those we are trying to affect change in to move product development forward. This way of practicing Data Science misses the mark because we are not listening to where folks want to go. We are throwing unintelligible complicated maps over a wall to travelers that have never taken a cartography class. They can’t make heads or tails of it and throw it away before heading into the wilderness. Is it any surprise when they get lost on their way to the top?
Data Scientists Need Empathy to be Impactful
Data insights only capture the potential to affect change. Potential only becomes change when you put insights into action. This can only be done if you have built trusting relationships with your stakeholders to ensure they understand you aren’t leading them astray. You are on this journey together and want the same outcome they do. Asking a stakeholder to read a report is relatively easy. However, asking stakeholders to understand what the report means, change their perspective and put the things they learn into practice is the real challenge and ultimate goal of Data Science. Trusting, empathetic relationships between Data Science and stakeholders are foundational in building the context and understanding necessary to get there.
We look at data to learn new things about our products. Why do we want to learn using data?
Changing Perspectives: Learning empowers us to change our minds in meaningful ways and update our mental models.
Teaching new skills: Learning enables us to achieve something we were not able to before.
Fostering Creativity: Learning expands our capacity for creative, impactful solutions.
Importantly, Data Scientists are trying to affect change in individuals who often have a less robust data background. Good Data Guides think deeply about this difference and proactively how stakeholders might receive our insights. Data Scientists that have empathy for the way others might interpret their findings are more impactful. Is the report written in a way that almost anyone could understand? What about their peers that have less context? Are they all using the same terminology in the same way (ex: what does retention mean)? Where is the executive summary? The one pager? The recommendations for next steps on the project? The meeting where we get together cross-functionally to discuss the latest findings? Where do they go to ask a follow up question?
Building Context and Fostering Learning
I spend much of my time thinking about how I might improve learning through increased empathy for those I am trying to guide. I don’t know what to teach them until I listen to them and understand the context of their tactics and goals. They might not know how to effectively use data to make a decision because they don’t have the deep data expertise I have cultivated in my career. It is through shared listening and learning that we achieve success.
Below are some of the activities I find helpful in being an empathetic Data Guide.
Pause and look for the broader context.
If product teams create a feature that encourages users to install a search widget is an installation metric indicative of our business objectives or is the true goal to increase downstream behaviors such as increased searches or ad clicks. Remembering the broader context ensures Data Scientists are guiding teams to hit our true targets and results are more impactful when they are couched in this context. Don’t understand the context? As a guide this might raise a flag and it is time to listen to where your stakeholders are trying to go and reconsidering what value your analysis provides.
Learn about data-adjacent roles.
Take classes that walk through the objectives of each role and expectations for what they need to accomplish in their day-to-day. This will help you understand what it is like to function in this space and what data insights are actionable for them. Make your recommendations specific to what is actionable.
In addition, learning about other roles gives you a deeper appreciation for other’s expertise and helps build shared mental models and vocabulary for deeper discussion. For example, I didn’t understand the processes and opinions developers were used to design features users love. In talking with our UX manager I learned one such framework is called the Double Diamond where designers start by exploring a space and gathering many ideas and then narrow them down to focus on what matters.
In learning about how this works in practice, I realized there is a step where Designers evaluate the cost/value trade off to narrow down their options anecdotally. This is a great place for collaboration where a Data Guide can provide quantitative Opportunity Sizing to evaluate which options we should not pursue because they won’t lead to the top.
Read business books.
Seriously! Data Science is conducted to support the growth of the business. Insights that are interesting but not impactful or actionable to the business objectives fall short of their goal. I recommend reading Escaping the Build Trap and Inspired. These business books discuss what roles organizations need to be successful and how they drive complex objectives forward.
Be aware of your educational differences.
Teaching and learning is an important part of connecting data outputs with product outcomes. I still remember the lightbulb moment I had when I learned what a regression is for the first time and how powerful that knowledge was. I wasn’t born with it, I have taken many statistics classes to hone my understanding. Help your stakeholders hone their understanding through delightful lightbulb moments with data and commend them for it. Correct them kindly and privately when they make an error so they do not feel ashamed of the mistake and reinforce best practices with everyone. Ask simple questions to your team publicly to model the behavior and demonstrate that no one knows everything.
Meet each stakeholder where they are.
This means Data Guides don’t stop when we encounter a stakeholder that doesn’t know SQL or can’t define retention for you. In fact, this when we lean in by creating forums with shared psychological safety where all questions are welcome. This policy is inclusive and builds strong relationships with everyone. Write a glossary of terms if you are introducing new concepts or your domain has its own jargon. Write at a level that you might use to talk about your work with your friend over dinner.
Be kind, opinionated and honest.
Data outcomes don’t always agree with the feelings and opinions of the stakeholders that want greater understanding. As Data Scientists we must deliver honest constructive feedback to guide better decision making. This is best received and builds a higher degree of trust when Data Guides deliver this feedback in a kind and clear way. I have found my work to be more impactful when I am not a center of ‘no’ or “right vs wrong”, but instead think of this moment as a chance to guide and reshape the product outcome in an impactful way. What part of their idea still makes sense (“yes, the data supports ____....”)? What part of their idea needs to change (“...and I recommend...for greater success in...”)?
Listen and discuss more, explain less.
Remember async communication, but create opportunities for synchronous communication to facilitate relationship building. I like to hold a biweekly meeting that is focused on collaboratively discussing our findings with stakeholders. I hold a separate meeting to discuss technical issues with my Data colleagues. This helps ensure our results are being interpreted correctly and influencing changes in our business tactics while giving data practitioners another space to talk data details.
Challenge yourself to be a more empathetic Data Guide by adopting some of these strategies in your own work and let me know how it changes the way you achieve success!
Thanks for reading people + products! Subscribe for free to receive new posts and support my work.