Are we facing an oversupply of data scientists?

Just the thought of the potential data points in those numbers is enough to make the average big data scientist salivate. But not Australia Post’s General Manager of Data Science, Silvio Georgio.

Recognised as the top data analytics leader of 2020 by the Institute of Analytics Professionals of Australia, Georgio highlights the importance of leadership, teams and stakeholders in his work.

“I feel privileged to work with talent who inspire me, leaders who present opportunities for us to rise to the challenge, and people who are passionate about the customers and communities we serve,” says Georgio.

With a skill set including a mix of soft skills, business skills and data science proficiency, Georgio has shown true leadership and empowered his team to transform Australia Post into a data-driven organisation. He is considered a Data Science Unicorn – a data science professional with both soft skills and technical skills.

But is Georgio as rare as his mythical spirit animal?

Until recently, it appeared that data science jobs were there for the taking – whether you had both technical and soft skills or not.

LinkedIn ranked data scientist as the single most promising job in the US in 2019 and it topped Glassdoor’s list of the best jobs in America for 3 years before that.

But what’s the job outlook for data scientists in Australia? Is there a shortage, or are we now facing an oversupply of data scientists?

We used big data and data science itself to crunch the numbers and determine whether the reality lives up to the hype. Here’s what you need to know about the job outlook for data scientist before stepping into the industry.

As a relatively new field of employment, data science doesn’t yet have its own classification in government labour force statistics. It’s usually lumped in with Actuaries, Mathematicians and Statisticians or Information Media and Telecommunications. That’s not going to help us paint a picture of the current landscape for data science graduates.

Luckily for us, even without raw data we can use data science to contrast how many job postings there are against how many graduate applicants. We can even specify the proportion of applicants from each field of study and whether they have a graduate certificate, bachelor or master’s degree.

To work out how many people are looking for data scientist jobs, we applied some old school statistical theory. We used Bayes’ law – an equation that tells us the probability of something, if we know the probability of a related event.

If we know how many university graduates there are, with Bayes’ law we can forecast the probability that they’ll be looking for work as a data scientist. That tells us the amount of supply we have for data scientists.

To use Bayes’ law, we need to know three things:

• the probability of being a data scientist
• the probability of having working in a field and level of education; and
• the probability of being a data scientist if you work in a certain field and level of education.

To determine the probability of being a data scientist, we turned to LinkedIn and searched for the number of people working in data science in Australia. LinkedIn returned about 26,000 results. It isn’t the most accurate number, but we can compare it against the Australian workforce to estimate the probability of being a data scientist.

We then used an automated internet browser to visit 300 publicly listed LinkedIn pages and scrape the data. While the data scientists we studied would never have known, surely they wouldn’t be surprised to hear that someone was hoovering up their data.

The information we scraped was simply the highest level of education for each individual, i.e. undergraduate degree or postgraduate, and the broad field of education of which they are an alumnus. This tells us the probability of being a data scientist if they work in a certain field and their level of education.

To complete Bayes’ law, we also need to know the probability of an individual having a specific field and level of education. Thankfully, we have a ready-made dataset from the Department of Education that breaks down the annual number of graduates in each field and level of education.

When each of these inputs was applied to Bayes’ law, we learned that the probable number of 2019 graduates who would be looking for work as data scientists was 5,000.

Now that we have an idea of supply, what does demand look like?

In October 2020 there were 700 data science job opportunities advertised out of a total of 150,000 job advertisements for the whole month. That monthly total of 150,000 was approximately the same as the average number of jobs for each month of the year. This means we can assume that there was an average of 700 data science jobs advertised each month – a total of 8,400 for the year.

With a potential 5,000 graduates applying for jobs in a market of 8,400, demand certainly hasn’t outstripped supply.

You’re probably thinking that all of those graduates wouldn’t necessarily be qualified for all of those jobs – and you’d be right. Add to that the fact that there are already information technology professionals in the workforce who would also be qualified for those roles.

So, we dug a little deeper. We gathered all of the data scientist job advertisements from Seek, Indeed, Jora and Careerone over a month. Each stream of data was structured and cleaned of duplicates so that we could work with it.

When we separated the junior roles from the senior roles, we found that 16.2 per cent of the data scientist jobs would be accessible to graduates. That’s only 1,300 jobs out of the total 5000 data scientist jobs for the year – or 3.8 graduates per job.

Then we did the same thing for software engineering. It turns out that there are 6,800 software engineering jobs accessible to graduates each year. But here’s the rub. There are 39,800 graduates trying to squeeze into those 6,800 jobs – or 5.6 graduates per job.

So, we can confidently say that the annual supply of data science graduates is not meeting demand, and the number of graduates competing for those jobs is lower than another field like software engineering.

Not long after the first computers made their way into the workplace, business management expert Robert Waterman declared that companies are “data rich and information poor”. Almost 40 years later, we know that 90 per cent of the world’s data was created in the last two years. In fact, you and I are creating about 1.7MB of data every second.

While this abundance of data provides businesses with impressive possibilities, capitalising and analysing them is easier said than done. In fact, many organisations are failing in their goals to become data-driven.

Industry needs more data scientists like Silvio Georgio at Australia Post. When he took over responsibility for data science and strategy, he could clearly see why the business was struggling to achieve its data-driven goals.

What Georgio found was a “top-down expectation that decisions would be anchored in data, and a belief in data’s potential – but the understanding of how it could tangibly improve business performance is less common.”

Georgio knew that his data science team needed better technical resources, but his first priority was to win over the hearts and minds of Australia Post employees – to get them to embrace data. Four years later, it sounds like all roles in the organisation are analytics roles.

“Everyone in the business – from boardrooms to the delivery truck – is willing to use data. Our focus at Australia Post has been to inspire the art of the possible with data intelligence, to take data to insight, and through to intervention.”

To achieve such a dramatic turnaround, Georgio specified three golden rules for Australia Post data:

1. Make it mind-blowing
2. Make it fun
3. Make it understandable.

Then, in addition to using data to improve business performance, he used it to improve the lives of employees. Doogie is a data tool that’s named after a classic TV show to make it fun. Georgio’s team developed the tool to collect and analyse data on the health and wellbeing of 35,000 employees.

“Our analysis identified correlations between non-evident causes of motorbike incidents amongst delivery officers, and found an incident is 50% more likely after a postie returns from a holiday of seven days or longer – regardless of how long they have been doing the round,” says Georgio.

While that kind of insight can help improve business performance and the bottom line, it’s a mind-blowing revelation for employees who have a right to a safe working environment. It also provides an understandable outcome that involves every worker in the data science process.

Whether you work in information technology, artificial intelligence, machine learning, business intelligence or a senior role in any other department, you’ve probably witnessed the mismatch between an organisation’s data-driven expectations and its data reality. As a data visualisation, Harvard Business Review suggests an inverted triangle – the wide end at the top is the C-suite’s expectation, while the point at the bottom is the data science team’s capability.

Having a postgraduate qualification like RMIT Online’s Graduate Certificate in Data Science can help business professionals gain critical data skills to put them in demand.

In a data-driven world, you will learn to apply a breadth of technical skills, including programming, analytics, data wrangling and visualisation. You will also expand your understanding of data and how to leverage it across sectors and industries. 

If you don’t have a technical background, RMIT Online’s Graduate Certificate in Data Science provides a broad basis of programming, analytics, data wrangling and visualisation. Then you can join those who do have a technical background to focus on using the language of data to become a liaison between data science teams, management, employees and other stakeholders.

Studying part-time and online without interrupting your current role, you can gain the skills to advocate for data up, down and across organisations.

Make your own data-driven decision to become a data science leader with a Graduate Certificate in Data Science.