As the former Head of Data Science at ANZ bank, and a former consultant, he shares some incredible insights, and offers valuable advice, to Data Scientists at any stage of their career.
You recently founded and launched the podcast Data Futurology. Can you please tell us about the podcast, the motivation behind it, and what you aim to achieve with it?
Yes. I’m excited about it. Over the years I have noticed that, in our profession, there is great focus on research, both as a career path and as a way to contribute to the industry, but there are not many other ways considered as a way to contribute. I think that’s partly due to the fact that we’re a young profession, and we’re born out of research, so it makes sense that we see research as the cool thing to work on.
However, my question is, can Data Scientists lead? Can we effect change in organisations such as business, government, startups or non for profit?
My answer is YES! And through Data Futurology, I want to help data scientists become and be those leaders, make those changes and have a positive impact.
Obviously, our work helps organisations to be more efficient and helps customers get more of what they want, but that’s only doable when there is leadership in those organisations that understand the power of Data Science. However, I think Data Science can, and should, be applied more broadly than what we’re doing currently.
Data Scientists can rise through the ranks, and be team leaders and managers, but I want to help data scientists to think about and be ready to lead functions, divisions, and organisations as a whole. This is so we can apply our analytical skills, data-driven decision-making and Data Science across everything in organisations. That is the way we’re going to have the maximum impact both as individuals and as an industry.
To do that, we need to develop a number of skills that will help us maximise our impact. Skills such as vision, persuasion and understanding of business. And we need to focus on practical delivery, on outcomes and on making a difference. We need to be relentless in pushing to see the results of our work make significant differences within our organisations. We also need to prioritise work that will make the most difference.
By thinking, building these skills and practicing Data Science in this way, I think data scientists can go on to lead functions such as marketing, finance, HR, or divisions or businesses, governments or not-for-profits. I believe that the analytical thinking and the use of data brought by Data Scientists, will help these organizations become much better and therefore the customers and consumers (us, everyone) will be much happier and we (data scientists) will be moving the world forward in the way it needs to go. Though Data Futurology, I want to bring this thinking to the foreground of data scientists and help them make these changes around them.
You were previously the Head of Data Science for ANZ Bank. What were some of the main challenges of the role, and what were your key Achievements?
It’s been tough times in banking the last few years. Major banks in Australia have been shredding 10% – 15% of their workforce every year due to a combination of competitive pressures and complacency created in the good times. During this period, I started in one of the major divisions of ANZ, which is about 5000 people and it’s responsible for over 50% of ANZ’s revenue. There was no Data Science team in this division and I was brought in to start the Data Science capability and figure out “what we could do with this data stuff”.
Over 3 and a half years, we built the largest team in the division, a team of 30 data scientists and software engineers that helped the bank differentiate itself in front of our customers and delivered tens of millions of dollars every year by acquiring new customers for the bank, helping cross sell existing customers and helping our customers (large companies) provide better products and services to their customers and operate more efficiently.
The first year was about educating the division, showing what was possible and coming up with our strategy. I did this through a number of prototypes and lots of speaking with people both to educate them on the potential and to understand where they were starting from. Then, the more successful prototypes delivered solutions that met stakeholders and customers “where they were” with regards to the analytics maturity. However, this doesn’t mean doing only and exactly what they ask for. Quite the opposite actually! Because, as a Data Science leader, you have to guide them through the journey starting at the analytics maturity level that they are in.
The second year was about building a team and becoming profitable. And this is what I’m most proud of: the team. I say that because having a great team, creating a fantastic culture, empowering the team, giving them a vision and putting them in front of the customer are all “leading indicators” of great work and a significant impact to the organisation. The second year was about executing our strategy in the best possible way, focusing on delivering value for our customers every week and staying agile to the changing demands and industry dynamics so we can keep creating work that makes a difference.
The third year was about scale. Today, this team of 30 is able to service over 10,000 business customers every month through 3 different product lines. The business customers include some of the biggest companies in the world such as Microsoft, Apple, McDonalds, Vodafone, etc. Each company and tens of managers within each client company use and benefit from the team’s analytics every day to build a better business and make their customers happier.
What are your views on successfully building and managing a Data
Start small, stay close to customers and deliver value every week, i.e: something customers can use.
I think there’s a tendency in our industry to do big projects or to think that we know what people want, or that things have to be “done perfectly” before release and I think that creates unnecessary barriers that slow down delivery and move our teams more into research rather than value delivery.
Secondly, I think that some people that go into management or want to go into management do so because they want to feel important, spend lots of time in meetings and/or get away from the code. These are the worst things you can do, in my view, if you want to build a successful Data Science team and career.
When you are a manager, the team doesn’t work for you; you work for the team. You have to wear 2 very different hats within the team. One is the person who gives them the goals to hit. In my view, these should be high level goals because it’s the team’s responsibility to decide how to get there. Once the goals are decided upon and agreed, you, as the manager, become a coach and a mentor to the team. This is where you spend most of your time within the team, and this is where you work for them. If they need something to be more effective or get the work done, it is your responsibility to develop the political sway to get it for them. If they have any roadblocks, you have to get rid of them for the team. This is what I mean that “you work for the team”.
Additionally, you have to be in contact with the different parts of the organisation and with your customers, but, your biggest asset is your team. The way you treat them will determine the teams’ success, the difference they make in the organisation and the way the team deals with customers.
What initially prompted your transition to Data Science? What do you enjoy most about it?
When I went to university there wasn’t much focus on data mining in our degrees, as you can imagine. In fact, I majored in Hardware Engineering and Embedded Systems, which is actually now coming back into focus. I only did a few subject on databases, algorithms and data structures and machine learning.
I noticed at the time though that every technology job I had while at university was in data: building databases, building software to capture data or analysing ways to run a business more effectively. So, when it was time for me to do my thesis, I thought I would look for a project where I could combine the 2 streams: data and hardware.
I was fortunate enough to find an industry-sponsored research project that aimed to detect tiredness in truck drivers in the mines. They needed someone to design the onboard “miniature computer” (as it was called at the time) so it could fit inside a baseball cap the driver would wear. They also needed help in analysing the data in order to detect and predict when a truck driver was too tired to keep driving. As you may have guessed by the baseball cap hint, the data we were using is from an electroencephalogram or EEG. Another person working on the project, a now retired electrical engineer, designed and built EEG sensors for a baseball cap by himself. Amazing.
In my contribution to the project, I designed the onboard computer to capture and process the data as well as communicating with the outside world via bluetooth to alert the truck and HQ about the tiredness levels of the drivers. A machine learning model would determine the tiredness level of the driver. I tried heaps of different approaches to do this and, luckily, we were able to get thousands of hours of labeled data, both EEG and video footage from sleep experts around the world. They had marked when the drivers in the footage were getting to certain levels of tiredness. That allowed me to make the model. I did build a neural network on my machine that ran pretty well with good accuracy but unfortunately (and maybe obviously), it could not run in the on-board computer. I ended up using a linear regression for that which worked pretty well too.
This project went on to become a commercial product which is still available today. That was my initial real world introduction to Data Science but, as you hint in your question, it wasn’t in high demand in industry back then. I then worked on Data Warehousing (what’s now called Data Processing Pipelines, and in Business Intelligence or Reporting (what’s now called Analytics). During that time, I noticed that people left a trace of their thoughts, actions and preferences in the data. Exploring that aspect more is what brought me back into Data Science.
I love being able to understand people at a deeper level through Data Science. I love being able to see their behaviours and preferences. I love being able to understand how they think what what’s important to them and then being able to provide them answers that they needed but maybe didn’t even know they needed.
I think that Data Science can have a huge impact if done right and that’s something that gets me really excited. I think the world needs the benefits that Data Science can bring and I’m passionate about helping this impact come into reality by helping others maximise their impact.
What areas of research are of particular interest to you?
My areas of interest are in the social sciences. As I mentioned, I got into Data Science to understand people and measure the impact of their decisions in order to improve decision making through data. For me, the areas of econometrics, psychology and marketing are really exciting, especially when they use common methods, in unconventional ways, to uncover insights previously out of reach.
For example, a couple of years ago, I read an econometrics paper that used Random Forest to understand the effects in people’s lives of graduating from university. This was interesting because, instead of using the straight predictions, they used the proximity matrix to understand the effects of graduating university. The proximity matrix is a calculation usually done as part of Random Forest but rarely used. With it, they found people that are “very similar” (i.e. close in the proximity matrix), but where one went to university and the other one didn’t, they then studied the effects of that decision in their life.
I have used that technique so many times in business, it’s crazy: from helping companies evaluate customers, stores, offices, mobile phone towers, etc. It’s extremely powerful, widely applicable and very beneficial in many domains.
What are some of the biggest challenges you’ve faced in your career and how have you overcome them?
Heaps of challenges, haha. Heaps and heaps. Almost 10 years ago now, I started an analytics consulting company with an old colleague of mine. We didn’t know anything about business and we nearly went bankrupt 4 or 5 times in the first year and a half. That period was tough. But then, we started getting clients, including some big clients such as Foxtel, MLC (part of NAB) and Government agencies. This meant a TON of work. For several months in a row, I remember billing over 100 hours per week, often 110 or 120, and that was “just” the client work, we were trying to build a business out of that so that needed time and focus too. That period was also very tough and, in both of these times, I made HEAPS of mistakes that made things harder on myself, but that’s how you learn.
When I started at ANZ, I had to learn about developing a data strategy from scratch. What you don’t see as a consultant (I was in consulting for 12 years before going into ANZ) is that there are many decisions, conversations and consensus required to get a project off the ground.
Generally, as consultants we came in towards the tail end of that process. At ANZ, I had to drive that from the start, when ideas are no more than seedlings, and help them grow in fruit bearing trees. It’s a long process and something that I found challenging because I enjoy more doing the work with the team, rather than talking about hypotheticals and trying to make others in the organisation feel happy, important and involved, even if they can’t really help. I had to learn to change. It’s still difficult, but it’s helped me in my career.
How have you seen the field of Data Science change throughout your career, and what advances, such as in the areas of AI and Deep Learning, excite you the most?
The field has changed significantly, especially in the applications of it within industry. When I started working 15 years ago, data was something you hoarded and migrated on system upgrades. Companies then started to get a few uses, but it was used sparingly. For example, I worked in telecommunications looking at the issues in the order to activation process (i.e. what happens in the background between you signing up for a mobile phone and when the mobile phone starts to work). Specifically, we were looking at the errors and aiming to fix/minimise them. Clearly, this project was more in the business intelligence/reporting area, which was the next wave of changes I saw in the industry.
Then, in the last 6 years, machine learning started to get more of a name for itself. Unfortunately, a lot of this has come from hype. People that don’t know about it read a magazine article about it and then they “have to have it”. This has helped create the awareness and demand in our field, but now it is time to transition out of it to a point where we’re making significant changes and contributions in organisations from a leadership perspective. Instead of waiting and applying Data Science where we’re asked or allowed, my idea is to change that entirely and have data scientists leading and applying Data Science everywhere.
I think deep learning is exciting and it will keep creating huge advancements in the world but then there will be the next innovation in ML. I think it will be around Bayesian networks or maybe a combination of Bayesian networks and genetic algorithms or neural networks.
What do you think are some of the biggest challenges limiting the uptake of analytics within organisations, in order to enable data driven decision making?
Lack of knowledge. By both Data Scientists and business people. Data Scientists must teach their knowledge to the highest number of people. We all need to be evangelists for our work. The business world is keen to learn but they find that our world is shrouded in mystery and they don’t know how to get in.
On the flip side, Data Scientists must also learn about business, finance, marketing, negotiation and the other things I mentioned above.
What are some tips you’d like to share with aspiring and junior Data Scientists looking to enter the field and develop their careers?
When deciding where to work, choose a great manager above all else. Don’t worry about the brand, the press, the pay, etc. Choose a great manager, someone who will challenge you, develop you and create a great team. Don’t worry about anything else.
Challenge yourself on soft skills and leadership from early in your career. Sign up to your local Toastmasters club and challenge yourself to at least finish the beginner training, hopefully stay for the advanced one too.
Work as a team. The most enjoyable part of Data Science is working as a closely knit group that has each other’s backs, overcomes huge obstacles and moves the dial in the organisation.
Think business. Learn business. Think strategically. Act strategically. Maximise the impact of your work. Maximise the impact of your team’s work. Measure your work by the value created and the feedback you get directly from customers. Don’t let anyone be a middleman between you and the customer.
Have a long term focus. Your career is a marathon. Work is about 3 things: acquiring knowledge, building skills and a network. Aim to learn the skills that will best serve you in the long term. Make sure you make time to build relationships that will make your work much more fun and satisfying.
Never stop coding. If your aim is to move into a role where you spend your entire time talking, doing meetings and being hands off the tools so you can feel important; then you have already failed.
Disclaimer: The opinions expressed by Felipe represent his own personal views and not those of his employer.