How Not to Explain Data Science to Business People
No surprise that Jarrod Teo’s job is all about predictions. But that doesn’t make Jarrod, Chief Data Scientist for DSS, in the least predictable himself.
Take for example a recent interview Jarrod gave to AI Time Journal. He told writer Nisha Arya Ahmed that jargon is a great way to lose the interest of business people, that the choice of software should not be the first consideration in a new data science project, and that creativity counts when it comes to hiring data scientists. Highlights from the interview:
Nisha: What unusual or absurd thing do you practice or advocate for in your profession as a data science leader?
Jarrod: This might seem unusual to some data scientists, but I don’t look first at what software we should use. I look first at the business objectives. When we start a data science project at Direct Sourcing Solutions, we use this sequence: Business objectives > Data we have > Data science methods to use > Software to consider.
Nisha: What advice would you give to business leaders who would like to step into realising data science use cases? What advice should they ignore?
Jarrod: Plan for the cloud but don’t use it until you are very convinced that the machine learning model is earning money. Start the machine learning model offline with a lower budget first because there will be experiments and trials with the data on hand to show you can solve the business objective. If you start with a cloud environment, costs can build up and undermine support for the project. So, the project must make money first offline and with a lower budget.
Ignore anyone who tells you machine learning models can be produced fast in any situation. I had a senior manager once who thought a new machine learning model could be created within seconds and without looking at the data. Machine learning models take time to build. Also, not all business objectives need machine learning models.
Nisha: What skills and attitudes do you look for when hiring data scientists?
Jarrod: As the Chief Data Scientist of DSS, I look for data scientists who have a statistics background and are creative. Yes, coding is important, but they can pick up software along the way. To be creative and able to read the output from the statistics software is very important.
Nisha: In your opinion, what have been the most relevant breakthroughs in data science impacting our world in the last 1-2 years, and what trends do you see emerging going forward?
Jarrod: AI Security. With AI now being used consistently, the topic of the moment is: What if our AI-trained models are hacked? Leveraging AI to identify cybersecurity attacks is a trend that can get more focus going forward.
Think about retail, for example. With the increasing prevalence of customised recommendations of products to customers, an attack on the AI pipeline can allow customised information to be obtained. Expect more interest in understanding how machine learning models can help to uncover patterns in cybersecurity attacks.
Nisha: What lessons have you learned getting companies to lead through data versus gut feeling?
Jarrod: Do not explain things using jargon. This is where the acceptance of outcomes from data will fail. Business people do not want their one-hour presentation on making money to become a statistics lecture. They only want to know how to make money.