The tech scene here in Charlotte is booming. In a few short years, the Queen City has emerged as a national leader in analytics and big data (and sometimes pro football). Here at RV, we’re working harder than ever to push the needle forward on the first two fronts. Weird coincidence, right? (Yeah, not really.)
We put data at the center of everything we do, which means we’re constantly learning and improving. It also means our data science team has uncovered some truly brilliant insights into this fast-growing, constantly-changing, impossible-to-predict space. And they’re sharing some of those with you.
Here are some data science best practices from the Red Ventures team — and your chance to adopt them* before 2018.
*the best practices, not the team
1. Organize Your Data
When it comes to organizing data, efficiency should always be your end goal. (We know, easier said than done.) To get there, you’ll need to make sure your data strategy reflects your business strategy. Since no two business strategies look the same, we can’t tell you exactly what your journey should look like. But we can tell you what we did. Our most important assets are our people. So, our first initiative was to merge our Data Team (the ones who produce the data) and Data Science Team (the ones who crunch the numbers). Because as they say: the team that works together, wins together.
2. Invest in Research
Totally official science fact: innovation requires constant experimentation.
Sorry, Bill. Not that kind of experiment.
To become a data science big shot, you’ve got to be bold. You’ve got to push the limits of what you know. Around here, we do that by constantly running different models and testing different algorithms. If we could, we’d run experiments every second of every day. But, as that’d be both super-expensive and against the constraints of the time-space continuum, we make the most of what we’ve got. How? By making our experiments cost effective. More on that in a minute.
3. Get Your Head in the Cloud
For most data-driven experiments, having better hardware (and less environmental maintenance) allows for quicker iteration. Quicker iteration leads to better learning.
When someone says training neural nets on a local 4 core CPU is the same as training on a beefed-up, GPU-supported AWS machine.
A cloud-based environment offers just that: faster, cheaper ways to test — and a higher probability of success. Given the crazy computing power and the minimal maintenance work required on cloud platforms, we think cloud will emerge as the next OS for data science.
Internet giants like Amazon, Google, and Microsoft all provide cloud services — and they make big bets in developing analytics and machine learning platforms. As big data becomes more accessible, you’ll see startups and small companies investing in these new platforms rather than new tech.
4. Buy Simple Analytics Solutions
When it comes to analytics solutions, you’ll come across this question pretty often: is it better to build ‘em or buy ‘em? Most out-of-the-box data science platforms provide advanced functionalities like clustering, classification, and regression. Sounds like one powerful toolbox, right? Yep. But, sometimes power isn’t all you need.
Too. Much. Power.
With a well-structured dataset and a clear sense of what to do with it, pre-made platforms produce useful results, fast. However, when you’re working with a complex business problem, diverse datasets, or a multifaceted goal in mind… it’s better to start with a blank slate.
5. Build Complex Solutions Yourself
Our tech team often faces challenges that require multiple rounds of strategizing, prototyping, and refinement before we’re ready for production. That’s because our data science solutions need to be integrated with our own tech stack, tailored to our overall business strategy, and maintained within our own expertise. Building analytics solutions in-house allows us to create custom-fit models that are integrated with our business strategy and technology stack from the start. With a solid “made-for-RV” foundation, we have control throughout the whole process — that makes it easier to augment operations for our different verticals.
6. Bend the Curve
Perhaps the most important (and enduring) trend in Data Science can be summed up through a single, elegant catchphrase: “bend the curve.” In case you aren’t aware of graphs and their intricacies, this concept is all about changing the conditions of a problem to improve the output over time. We’ll attempt to illustrate with a gif:
Whoa, trending up!
At Red Ventures, we’re constantly re-thinking our inputs. It’s how we avoid complacency. It’s why we’re able to bring advanced digital platforms to many different industries. Most importantly, it opens the door to a whole new world of data. And you know what you get with more inputs? A higher probability for successful outputs.
Did you already know everything you read? Okay, genius. We like your style. Learn something new about our award-winning data science team and then check out tech positions we’re hiring for RIGHT NOW. (…unless you already know about those, too. In that case, go ahead and add “ESP” to the skills section of your resume.)