You Can’t Buy Your Way Out of Data Science Problems You Can’t Buy Your Way Out of Data Science Problems
You launched a data science initiative, and the results have been underwhelming. As a business leader, you understand the value of... You Can’t Buy Your Way Out of Data Science Problems

You launched a data science initiative, and the results have been underwhelming. As a business leader, you understand the value of investing in a project with a high potential ROI, but the project just siphons funds without showing improvement. So what’s the issue? Data science problems aren’t like other business issues. Underfunding can certainly kill a data science initiative, but that may not always be the real issue. Turns out, you can’t buy your way out of your data science problems. Let’s take a look at a few reasons why.

[Related Article: ODSC East 2019: Hilary Mason on Data Science Product Design Problems]

Your Data Isn’t Good

Garbage in, garbage out. It doesn’t matter how much you’ve spent on your data lake or your newest data engineer. If your initial data isn’t right for the project at hand, your results aren’t going to be up to par. 

All the funding in the world for state of the art tech or the newest shiny thing isn’t going to help unless you get your fundamentals right. Re-examine your data carefully, so you know that you’ve got what you need.

Instead of wasting money, do this: 

Meet with your data scientist (or team) and your relevant stakeholders. Discuss the data you have and the data you need. Figure out how to make those the same thing. It could be missing data – do you have the right emails? It could be the wrong kind of data – you have a geographic location, but you need demographics. It could be data mistakes – duplicate entries. Once you’ve gotten the data you actually need for your project, you can get started.

You Aren’t Using the Right Programming

We’ve talked about when to use deep learning and when to stick with machine learning before, but businesses are still making this mistake. Shareholders in board meetings get ahold of buzzwords, and you’re left figuring out how to work deep learning into your pipeline.

Machine learning is great for structured data because deep learning is too time-intensive for that. If your data is largely unstructured, ML will be underwhelming. It may feel frustrating to unravel what the right answer is, but once you do, your projects will get off to a better start. 

Instead of funding, do this:

Get with your data team and figure out what kind of information you’re working with. Create a plan of action you can take back to your shareholders to outline what type of programming you’ll work with and why.

Wrong Frameworks

You’re going to need the input of your data science team on this one. Not all frameworks are the same and simply buying into the most expensive one, the one with the most “stuff,” the newest, shiniest thing won’t help. 

If you trust your people to do what you hired them to do, they should have a clear idea of the right framework for your investment. Also, there’s no such thing as a one and done solution. You’ll need a clear understanding of what tools fit your purpose. 

Instead of wasting money, do this:

Consult your data science team about what works for them. Look for ways to break down silos between your data science team and other stakeholders. Think business analysts, finance, or sales teams. If you’re Agile, consider how new frameworks fit into your pipeline.

You Don’t Have the Right People

Same story over and over. You send out calls for a data scientist, but what you really need is a data analyst. HR pores over resumes for a data scientist, but you need a machine learning engineer. Data scientist seems to be the catchall term, but in reality, your job title matters.

It’s not that people can’t find your add. People in data science are sifting through multiple searches to find jobs that align with their skills. However, if you’re clear about what it is you need, you’ll spend less time with irrelevant resumes and more time finding the talent you need.

Instead of wasting money, do this:

Put in some research on data science job titles and get clear on what you really need for your problem. If you still aren’t sure, get really clear on what your first business outcome is and go from there in your job posting.

Money’s the Answer… Until It’s Not

Data science isn’t a cure-all. It’s not going to turn around a failing business if your business leadership is lacking. It’s not going to give you insights if you don’t know what question to ask. Funding is a critical part of any business initiative, but don’t be afraid to slow down and consider what it is you’re buying when you approve new expenses. 

The right person with the right frameworks and the correct data can do more for you than all the fancy, shiny new tech out there. As always, start with your big “why” and form your data science initiative based on that specific why. You’ll get further and save your spending for where it’s really needed.

Elizabeth Wallace, ODSC

Elizabeth is a Nashville-based freelance writer with a soft spot for startups. She spent 13 years teaching language in higher ed and now helps startups and other organizations explain - clearly - what it is they do. Connect with her on LinkedIn here: https://www.linkedin.com/in/elizabethawallace/