Top 5 Data Science Pitfalls And How To Solve Them

Scott Plewes | October 10, 2019 | 5 Min Read

Avoiding these pitfalls will help you focus your efforts on leveraging data science as a tool to solve real business problems and avoid undertaking projects for the sake of being able to say you’re leveraging data science.

Often times in our discussions with medical device vendors that have attempted internally, or with another consultancy, to apply data science methods to the data they’ve worked so hard to enable their device to collect, we’re told that their data project failed to live up to expectations.

To date, many data science initiatives that medical device vendors have undertaken have either outright failed, ended in a big question mark, or stopped at the Proof of Concept stage. Many vendors will go the distance to enable their device to collect loads of great data and then the team is unsure of what to do with it and how to use it from there.

One of the lessons from this insight is that data science opportunities are just like other product or business opportunities. Whatever your specific design and development process may be, you will still go through an analogous discovery, design and implementation approach that has worked with products that didn’t utilize data. In the end, your data science-enabled product still has to meet user and market needs.

The result of a pitfall often boils down to your product team not realizing business value. You could do “technically excellent” data science, but if there’s no ROI, you’ve fallen prey to some issue.

Here is a list of the top 5 data science pitfalls we see and our suggested solutions.

Pitfall 1: Underestimating the Data Engineering Required

You can’t do data science without data – specifically, “good” data that feeds the data model or algorithm you are using to gain insights or make predictions. The effort involved in getting data in a form that’s useful for analysis is often underestimated by companies and organizations.

Solution:

There are a few things here that are related. One, your product team needs to figure out what data you actually need, which requires developing a data strategy that maps to your product and business strategy. Second, ensure your team scopes the data engineering effort properly – it’s commonly underestimated when starting a data science initiative. And finally, it needs to be understood that a data engineer typically has different skill sets than a data scientist. You need the former to properly understand the data engineering required, not the latter.

Pitfall 2: Chasing the Data Science Hype and Expecting Miracles

It’s hard to go through a day without hearing about AI and data science in the medical devices or technology world. Without a doubt, there are some amazing things being done and even more amazing things being predicted. However, these are susceptible to the hype curve just like every other relatively new idea or technology.

Solution:

Don’t get caught in doing data science for the sake of doing data science. Make it an instrument of change, not the end goal. If you’re succeeding in all the other aspects of the business, stand by your business model and processes to identify IF and how data science can be incorporated to add value. Data science has a role – sometimes a critical one – in strategy, planning and execution, but it shouldn’t drive these things any more than technology or project management should drive the business.

Pitfall 3: I Just Need to Gather More Data

With all the excitement around AI and data science as we’ve just discussed, it’s understandable that every time an issue or potential issue surfaces, product teams believe more data is the solution. While it’s true that sometimes you will discover enlightening patterns by just “looking” (mathematically) at data, it’s a costly endeavor if you don’t know what you are looking for.

Solution:

Not all data is created equal. Focus on the questions you need to answer first, as this will drive the data you need to gather. Deciding on a data strategy requires you to make the connection between the data you’re going to gather and your business goals. In the end, the effort you put into gathering data, as well as formatting it correctly and getting rid of “garbage” data that doesn’t serve your business goals, will be a reflection of both how hard that is to do, and how valuable it might be. Your team will identify data that is mission-critical to your business goals, and thus, is worth the time and energy to collect and sort.

Pitfall 4: Everyone is Speaking a Different Language

Like engineering, design, product management and so on, data scientists have their own lexicon. Quite often, terminology and concepts are either foreign or actually contradictory to other members of the team. A “feature” to a data scientist is essentially a variable, but to a product manager, of course, a feature is a characteristic of a product. This sort of problem can lead to teams getting out of sync.

Solution:

Data scientists need a minimum ability to converse in project management, product, technology talk, etc. Similarly, the other disciplines need some “data science 101”. It’s also a huge help to integrate teams as early as possible in a project, even if some teams’ participation is minimal.

Pitfall 5: The Business Goal Wasn’t Clear

Typically there are two ways to start a data science project. One is to start with a business problem/goal and use data science to solve it. The other is to explore the data and uncover a business opportunity. In both cases, at some point (sooner rather than later in general), if the business goal isn’t clear than the data science project will fail.

Solution:

The solution to this pitfall is exactly like any other project in this regard. Defining clear goals and knowing what success looks like when you get there is essential. Don’t get caught up in solving only a data science problem.

Data science is an incredible – and still nowhere near fully utilized – tool that has the potential to improve care delivery on every dimension. For your efforts to leverage data science to succeed, it will be imperative to avoid the pitfalls we’ve mentioned above. Hopefully, this advice will help you focus your efforts on leveraging data science as a tool to solve real business problems and avoid undertaking projects for the sake of being able to say you’re leveraging data science.

Download: Applications of Data Science in Medical Devices

In this eBook we demystify the domain of data science, articulate what’s involved, dive into the two main approaches to applying data science, and review the most common pitfalls experienced by teams that lead to failed data science initiatives.

Suggested Stories

Health Information System Integration

In this webinar, we discuss interoperability in healthcare and answer attendee questions on Health Information System integration. Download the webinar Now.

Read More

Achieving Interoperability in your Healthcare Organization

The challenge with integrating healthcare organizations is that they are just that – separate organizations. Each organization provides a healthcare service, but they provide that service using their own process.

Read More

Using Contextual Data in Connected Device Design

With more physical things being connected to the internet, the amount of data collected by these devices continues to grow. When designing an experience for an IoT device, considering context becomes increasingly important to avoid agitating users.

Read More