At my last job, I was a data science leader and managed people for the first time. I did a few things right, but a lot of things wrong. I recall one seminal moment, when my leader gave me my last performance review before leaving. It wasn’t exactly negative, but somewhere south of favorable. He indicated (and I’m paraphrasing from memory), ‘Lindsey, I have one simple metric for performance. If you added value to the company, you performed well. If you didn’t add value to the company, you didn’t perform well.’ It’s a moment I’ll probably always remember and still think about often. He’s a wonderful leader and I learned a lot from his mentorship. In the months that followed, it seemed like a lot of data science managers at other companies were being asked to do the same thing—measure their impact/value to the company. I appreciated the simplistic rubric, and data science value to a company is still an evolving topic for me. But as of now, I think measuring the value of data science and analytics to a company is a complex thing that doesn’t follow some y=mx+b line or some binary threshold. In my opinion, it’s a lot of hand waving to tie a machine learning model to revenue/sales cycle unless you are actually selling the model itself. And it doesn’t really help shareholders or augment company strategy to handwave. If you (really) need machine learning or statistics model to solve a business problem, then it doesn’t matter if you have 1 client or 10k clients. You invest in it anyway. 

 

Which brings me to the point of this blog post—what do the clients and business side think about data science output and its value? Do they measure it with emotion? As a trained scientist, I was taught that science is necessarily dry. It exists to advance knowledge about our physical and chemical world and results should not be open to wide interpretation. You have a problem, you design and experiment, you execute on the experiment, you reject or accept the null hypothesis. The End. On to the next iteration. There’s no feeling. There’s no crying in science! I don’t feel anything about the results or data—they exist absence of any feeling I have. Same for machine learning and regression I might create at a company to solve a problem. I have the data, I create a model, the model has performance scores that are either acceptable or not, and that’s it. I don’t feel anything about the results—they are dry and have distinct values with some confidence interval. Sure, you can wax poetic over the results with a few beers on board. But that doesn't change the dry results.  

 

But, as I’ve learned over the past few years, there’s a lot of folks out there in the business who are having a lot of feels about data and output, and rightly so. I read about Scott Galloway’s T Algorithm recently, an 8-item quality list for creating a trillion dollar company. The first item? Appealing to human instinct. Interestingly, ‘visionary storytelling’ is also on the list, which piqued my interested due to the focus on data science storytelling in recent years. When I learn things like this from economists and business leaders, I can’t help but try to tie these concepts to data science. Are we thinking about the emotion involved in the outputs? What are users doing with these numbers we create, and how should they feel about it, if anything? If the number changes, does that cause angst or worry? We might think those changes make perfect sense from a science perspective, but don’t translate well to human instinct. This is especially true in the healthcare sector, where numbers and values and graphs and such are being used to make healthcare decisions, move people to different facilities, intervene when a poor outcome is likely to happen, etc. 

 

I propose that data scientists should start thinking about the human instinct aspect of data and outputs. The mathematical explanations don’t always seamlessly translate to human instinct. Sure, it might be nothing more than an optics issue at times, but it’s important to remember that our science that along with iterative knowledge, is also being used by real people in real time out in the wild for something. Smokey Robinson would second this emotion.