Lessons to avoid "failure" in Data Science projects

During a live session on "When Data Science projects fail" by Dr. Jacqueline Nolis where she picked some of her notable challenging projects and explained the barriers faced by her, I learnt the following key lessons:

Lesson 1: Poor implementation leads to a failed project

It is said that first step towards a goal is the most important thing. But with this thought, we tend to ignore the relevance of proper "Planning and Strategizing". Hence, it is crucial to enforce an idea or a decision in an outlined manner to achieve the end goal effectively.

What to do? When starting a Data Science project always chart out a clear roadmap and set small attainable milestones. This will give a realistic shape to your idea and increases its probability to success.

Lesson 2: Uncertain objective leads to a failed project

A best decision is taken only when the exact requirement is well-understood. A single step taken towards the right direction is always better than 100 steps taken towards the wrong direction.

What to do? Prior to project planning, always discuss the clear objective of the project with the stakeholders to build a meaningful project.

Lesson 3: Poor leadership leads to a failed project

A project is accomplished with the efforts of dedicated individuals from various teams with unique skill sets. And a good leader is needed to synchronize the expectations of all the team members and set an equal bar for all of them.

When more than one team comes together for a project, a 'collaborative attitude' should be promoted instead of a competitive attitude.

What to do? Have a realistic expectations and a transparent communication amongst the team members.

Lesson 4: Building inaccurate models are normal

Failure teaches you what success can not! Building not accurate models at initial stages is okay. Taking risks and exploring new techniques to tune the model is a good practice. Any successful innovative idea has gone through unwaveringly high number of failed tests.

What to do? Accept failure with a positive attitude and embrace new learnings the tough way. As Thomas Edison said "I have not failed. I have just found 10,000 ways that won’t work."

Lesson 5: Customer needs is paramount

The decision lies between what is required and what is accurate. If a customer is satisfied with 75% of accuracy until he is getting the correct figures to take the business ahead with optimal finances, there is no need to invest sweat and blood to build a model with +90% accuracy. Accuracy is at the discourse of the customer and its business requirement.

What to do? Follow the golden rule "Customer is First"

Hope these principles by Dr. Jacqueline Nolis help you to align your thoughts on how to effectively approach a Data Science project, like it helped me.