Insights

Five Reasons Data Science Projects Fail

Analytics & Insights

Authors: Brian Liberatore and Jeff Renz

Data science is at the core of the innovations that have shaped the past decade – from self-driving cars to social media marketing. It is fueling a powerful new class of analysis tools that are transforming how we operate and understand business. Data science also routinely drains millions in high-profile forays into machine learning that leave little to show.

The difference between a transformational project and a resource drain lies not in the algorithms, but in the integration. Success in this field requires more than data. It requires a holistic approach that integrates data science into an organization’s operations while anchoring directly to its top priorities.

Data Science works when its practitioners prioritize people and processes – not just data and technology. Solutions must work with your business, not apart from it. This is vital to assuring a project’s value justifies the investment.

 

Data Science Venn Diagram

5 reasons data science fails to deliver

Data science projects present unique pitfalls. Recognizing them is key to preventing them.

1. Siloed approach.

Data science projects cross many disciplines – statistics, computer science, business operations, application development, and more. Projects and practitioners need to see the entire picture and cross between disciplines – something many organizations struggle to do.

Breaking from the silos requires a framework that accounts for all aspects of the organization. One must understand what drives your business – its people, its processes, and its technology – to effectively harness the power of data science.

2. Answering the wrong question.

Projects that start with the question, “What can we do with this data?” end up answering a question at great expense that often doesn’t matter to the organization. “Why do we do this?” is a better question than “How do we do this?”

There needs to be a clear objective and an impact you can measure. A project is successful only when you can describe its impact in concrete terms.

3. Failure to integrate.

Solutions need to integrate into a business’s operation to have an impact. Otherwise the insights are lost. Business leaders can’t make decisions based on information they don’t have. Projects need to make that transition from pilot to production. According to Gartner, a global research firm, most projects never cross this divide.

Projects must also work within an organization’s existing communication channels, its people, and its existing technology. This way the right people are seeing the insights and can act on them in the right moment.

4. Stakeholders disengaged.

Some project management approaches leave stakeholders in the dark on a project’s value until the end, when it’s too late. Disengagement sees projects drift from the original goals. Feedback that comes too late can’t steer a project.

Agile project management is a good antidote to this. The approach integrates stakeholder feedback early and often, which means the project’s value is always clear and to can adjust to stay focused on delivering value.

5. Benefits don’t justify the costs.

Just because an organization can answer something with data science doesn’t mean it should. Results need to measurably exceed the investment into data science – they often don’t.

The best tool is often the simplest one. There is no need to build a custom algorithm if there are tools that can get the job done just as effectively, but faster and cheaper.

 

 

Brian Liberatore has delivered data science projects around the globe, always with an eye on the projects’ value to the client.

Jeff Renz has more than 20 years’ experience creating high quality data warehouse and analytical solutions across several industries.

Subscribe to our Newsletter

Please Check to Accept Our Privacy Policy