The Five Things to Consider for Successful Data Science

Truly successful data science projects have five things in common…

A successful project is implemented and generates true business value. Yet such data science projects are unfortunately still rare.
Projects fail mostly for the same reasons, turning these reasons for failure around create the five components for success in data science.

man sitting on mountain cliff facing white clouds rising one hand at golden hour
Photo by Ian Stauffer on Unsplash

Let me introduce you to them…

1. Data & Tools

Your data and tools need to be data science ready. Companies still struggle with getting access to the data (both politically and technically). Many companies have legacy systems that make the lives of data scientists hard. Not understanding the requirements or the different possibilities of tools used for data science or not having the funds available to give the team what they need to work efficiently and effectively. More than once have I heard people taking home their models on their private GPU because the company infrastructure wasn’t sufficient.

Does your team have the necessary data and tools available to do the work?

2. Skills

You may think of the skill to build a deep learning model, data and software engineering, building a pipeline. But that’s not enough. The skill to understand the data that is used in their business context and the ability to understand the connections of different data-fields is just as important. And let’s not forget the project management or product management skills, seeing the big picture and the ability to communicate to the business.

Does your team truly have all the skills available it requires to do the work?

3. Communication

I have mentioned this in other posts: communicate, communicate, communicate. Often conversations happening between data scientists and business people occur on two levels, they speak to each other and are under the false impression that they understand the same thing, the Analytics translator is the solution. Teams tend to jump into action without properly clarifying expectations from different stakeholders in the business. And let’s not forget the change management towards the end-users, especially if they are scared to lose their job.

Is your team communicating enough with the different stakeholders?

4. Business Case

The business case, a classical one when you speak about projects in general. But less classic in data science projects. Too often data science projects are started without a business case but through “Wouldn’t it be cool if…?” (fill out your favorite neural network and random application) questions. Avoiding an unclear project scope or checking how much it would cost to run a model in production and build a product around it can save you a huge amount of dollars.

Is your business case for your data science project clear?

5. Governance

If you have the data, tools, skills, communication and business case. The governance can still throw you back. The data science team needs to be embedded in the right place in the organization and have the right connections to the different supporting functions. The process around accepting projects, the way the team works and complying with rules and regulations are key. And last but not least you require the buy-in from the top management. If the governance is not set-up right, you are set up for failure.

Is your governance setting you up for success?


If you can ask yourself about your data & tools, the skills in your team, how you communicate with the rest of the business, the business case, and governance setup, you can quickly clarify if you are set up for success in your data science projects.

About me: I am an Analytics Consultant and Director of Studies for “AI Management” at a local business school. I am on a mission to make data scientists happy (again) and to help organizations generating business value with AI.