DATA SCIENCE: FINDING A COMMON LANGUAGE TO SUCCEED IN MACHINE LEARNING

June 05, 2019

By: Jared Rodecker, VP, Advanced Analytic Solutions, RAPP LA

The complexity of machine learning algorithms and their recent rise to market prominence can create challenges for many companies — just not in the ways you might think.

Seeking automated tools or the ability to make more evidence-based decisions, business owners bring data scientists on staff. A logical decision, but once embedded into the organization and aligned with marketing, frustrations can develop and data science challenges can arise. Few people speak the same complicated language of math, statistics, and coding as data scientists. And without intervention, frustrations can lead to friction between teams, which will inevitably hold back business.

Consider the last project your organization tackled. Chances are, more than one department was involved. If the different disciplines couldn’t find a common vernacular, that project probably would still be limping along to this very day. Communication is necessary to collaboration and vital to setting clear expectations and parameters companywide.

For data scientists, language barriers can be even more problematic, due in no small part to the role that data now plays in business. Marketing and sales teams have clear business objectives and KPIs but may not be versed in expressing these objectives in the mathematical language that data scientists would use to bake these objectives into machine learning algorithms. Though both groups are working toward a common business objective, it can be difficult to get both parties on the same page. Addressing the challenges of data science is tricky, but it’s not impossible.

The Benefits of Interactive Web Apps

Bridging the gap between disciplines, especially when it comes to data science, often takes one of two forms. Some companies prefer to cross-train divisions so that certain teams learn about data science and the data scientists learn more about business. Other companies place a variety of resources within the departments, creating different dotted-line reporting relationships than you’d find on a “traditional” team.

While both approaches can be effective business practices that build connectivity over time, their application is top-down and the benefits to the organization take time to accrue. Web-based applications, however, provide a bottom-up opportunity to jump-start the process and gain immediate, long-lasting traction before the larger goal of full organizational integration is achieved.

Consider a sales team that needs to forecast lead volumes in the coming months in order to ensure proper staffing levels for handling the flow. The benefits of an interactive web app that embeds forecasting models developed by the data science team are plenty.

Primarily, they can help improve efficiencies within an organization. The sales teams can cycle through the equivalent of 100 back-and-forths with a data scientist in mere minutes to forecast lead volumes under different market conditions, and the app can make it feel as if you have your own personal data scientist on staff. Think about the solutions you could develop by leveraging the knowledge extracted from data, which is set to reach 44 trillion gigabytes by 2020. That’s a lot of power just a click away.

Bridging the gap is sometimes easier said than done. But to create a better system of feedback and to bring data science into the fold, the following are often the best places to start:

1. Prepare employees. Sounds more foreboding than it actually is, but you do want to prepare your company prior to introducing data scientists into the business. Look at the organizational chart and decide whether you want an integrated structure in which data science joins the rest of IT or a specialized department in which all operations around machine learning work independently.
   
   Apart from that, you may also want to consider training requirements: not only for data scientists to learn about the organization, business, etc., but also for staff members who are most likely to work with this team. Existing employees should understand your data strategy.
   
   
2. Reevaluate expectations. Project management will look a bit different with data scientists on the team. You can’t just break a project into smaller, more palatable pieces and then set a time frame by which to monitor its progress. When data scientists set out to build a machine-learning model, you’ll be witness to some trial and error. As such, plan for quick iterations instead of a more measured pace toward a goal.
   
   
3. Work toward full integration. As soon as a viable app-based project can be identified, bring together business and data science. Encourage the divisions to sprint through a project and build an app side-by-side. Exchanging ideas and working so closely will help strengthen those bonds between teams and speed up the process of fully integrating data science into the business.

Data science challenges exist because the discipline works differently from others, but this dedicated group should be treated like any other team. Sure, you’ll experience some growing pains — it happens when introducing anyone into an organization. But if you prepare for its introduction, set expectations, and encourage collaboration, everything should go smoothly.