Cover image for πŸ€·β€β™€οΈ Why Do Data Science Projects Fail!

πŸ€·β€β™€οΈ Why Do Data Science Projects Fail!

waylonwalker profile image Waylon Walker ・1 min read

career questions for data folks (10 Part Series)

1) What is YOUR advice to NEW Data Scientists. 2) How did YOU enter Data Science 3 ... 8 3) Why do YOU Data Science 4) What is Your Advice for a new Team Lead 5) How do YOU debug 6) Variables names don't need their type 7) What is YOUR MVP 8) πŸ€·β€β™€οΈ What are some misconceptions about data science?? 9) πŸ€·β€β™€οΈ Why Do Data Science Projects Fail! 10) πŸ€·β€β™€οΈ Which DataScience related blogs/ magazines. Do you read? ??

What causes data science projects to fail?


markdown guide

Well, Waylon, I think it's due to many reasons:

  • Most people take data science as a complete solution to their problem however it's not like that. It might help you to a certain extent in improving your product but still, there are many things other than data science that need to be considered too while thinking of the big picture.
  • Data Science is a tool like any other tool. You as a decision-maker need to question yourself that is the problem under observation a data science problem or some other type of cognitive computing problem?
  • While starting a data science department at a company, most companies end up hiring a lot of people and start applying data science to their every solution. It's risky too and your team might end up in doing nothing. Instead, there should be an iterative and gradual approach to the application of data science. If we follow steps of a data science process and then iteratively improve our approaches based on the response from stakeholders then it might be a better approach.
  • Sometimes companies that start new data science teams end up incorporating people hired as data scientists to other teams like Business Intelligence, Database Administration, and other tasks due to the lack of proper communication of data science insights and value it can bring to the company to upper management.
  • Most of the times problem with us is that we don't understand as a data scientist that where to start from. So, at that point, we should prioritize things that are related to KPIs or which are key factors in bringing revenue to the company. It's necessary for the beginning because that way you can provide some monetary value to the company which might open up new chances of experimentation for you in other departments of the organization too.
  • Mostly failure is also due to the inability to reason for the use or working of complex models. For that, in the beginning one should always prefer the simplest models that are easy to the reason for and can be explained easily.
  • When presenting data science results, try to relate them to the value in terms of money this data science process is supposed to bring to the company. Instead of that if you get very good insights after a process, and it brings no value to your company then these insights might be helpful for a research student but not for a company in business terms.
  • Knowing your tools is necessary. You should know the basic ins and outs of different languages and frameworks involved in the data science process. Also, try to upskill new people by different courses or maybe meetings so that whole team is at the same point in decision making and implementation.