9 Mistakes to avoid when investing in Data Science

1. AI is magic

AI is a set of algorithms that learn from data using more or less sophisticated statistical techniques; It is not a spell that will solve your problems as if by magic – if anything it will, more likely, initially create more problems than it solves. Organisations should have realistic expectations and reasonable roadmaps to ensure the success of AI projects.

2. AI is IT

AI is not IT – it’s mostly business and how to change it! To bring the transformations needed, a Chief Data Scientist (CDS) should have a disproportionate power and be at the management board of the organisation and be backed directly by the CEO. AI is not IT because it involves much more than technological execution – it’s about making a data driven company.

3 Top management ignorant of Data Science

If managers don’t understand the basics of data science they will not put insights into production and the opportunities will be lost. In particular, in hiring and retaining talent will be highly ineffective as they can distinguish between BS and real talent. They don’t need to have a PhD in Statistics, but they need to understand the fundamentals.

4. Underestimate the skills required

A data science team requires much more than software engineering skills. It requires good knowledge of Statistics, Mathematics, and a long list of coding skills and methods like Python, SQL, Tensorflow, Keras, XGBoost, Could computing to make production ready and scalable code that can have a real impact in business.

5. Underestimating the challenges necessary

A data science project is very hard with many possible bottlenecks – lack of data, poor data quality, lack of resources, bad infrastructure, undefined goals, complex procurement processes, legal and privacy issues, etc. Be prepared for the long and painful journey.

6. Not being focused on value and measuring it properly

Data Science boils everything into numbers. So, it’s a good idea to measure every stage in order to remain focused. However, don’t be obsessed with measuring everything very often or just on going after easy cost savings projects – the so called cosmetic data science. Data science to be impactful has to be bold and disruptive (normally painful) – in this case the goals will not be easy to quantify. But you need to support your claims with testable hypothesis.

7. Thinking short-term

DS requires a long-term strategy. Even if it always wise to have a balance between quick wins and long-term moon-shot projects.

8. Ignore communication

Data Scientist are seen as introverted geeks that prefer watching endless lines of code incarcerated in a cave then talking normally like other humans. However communication is key in delivering a successful data science project.

9. Data Science is not Science

This is probably the biggest mistake. Even if Data Science is a prostituted word, it is more about Science than anything. It’s treating your company as a gigantic lab, always testing hypothesis, validating results, refining models, creating new ways of improving products and serving clients, challenging the status-quo.