Why Data Science Projects Fail ?

Many places I keep observing only failed #ML implementation. Data Science is here to stay but there will be a bubble burst with "Failed Data Science Projects"

Some of the lessons missed I observed are

• There is no single model to solve everything - Detection, Classification, Scoring, Recommendations. It will be a mix of multiple models
• Data Science project is creative work, needs data, finetuning, re-training, come out of deadlines, it's making someone learn from data
• The perfect model comes after iterations, not in the first iteration
• Don't go for AWS, Google, Microsoft Vision, NLP, AI tools in the first go, It is pay per use. Good to start. Build something on your own, You will control on internals and improving it further
• Production deployment - There are tons of tools out there, Building is more important than deploying
• Data Science projects and Sales teams visions often conflict. I have seen Sales teams not able to sell AI products because they don't understand fully how AI can fit in the portfolio
• Data Science is multiple perspectives = Data + Domain Knowledge + BI + Computer Vision + .... It's not just .fit and .predict. Learn the complete perspective. Don't get carried away with one perspective.

Happy Mastering Data Science!!!