Having the word scientist in your job title is surely quite a feat. Despite the instant whiff of intelligence associated with such expertise, next comes the notions of discovery, experimentation, 'world firsts', a splash of power perhaps, and of course, the stigma of a zany goggle-clad genesis.
Stick the word data in front of it and it becomes less about test tubes and basements and more about numbers. This may sound somewhat less sexy, but being a data scientist in the 'ad industry' surely takes it back up a notch right? Either way there seems to be a trend of data scientists bubbling up throughout ad land terrain.
From a glance at LinkedIn it's clear to see such roles are in vogue, with the likes of Betfair, Ninemsn, Optus, and Coca-Cola Amatil currently looking for data scientists – for some it's a new unicorn (Slingshot just announced its first), for others they are just stocking the stable with more.
Looking at one job application for a data scientist consultant at Teradata in NSW or Victoria, there's already been nearly 200 applicants. UTS is also kicking off an Innovative Data Science course this year so one thing's for sure – there's a hunger for it and there's clearly a need.
Who are these data scientists? What's the science behind it and where is the data extracted from? Stand guard as we have the answers from data scientists themselves hailing from the likes of News Corp, Optus, Vodafone, Red Planet, Slingshot and Mi9.
Tai Elliott, data scientist at News Corp Australia:
Like most jobs my job has a sexy side and a less sexy but very important side.
The sexy side is using technology and analytical processes to improve our understanding of what our users are interested in and when they're most interested. The most interesting advance is in an area called machine learning: teaching computers to understand stories written by our journalists and how our users consume them. The end result enables us to create compelling new advertising products, improve editorial insight and provide a better user experience.
For any of the interesting work to happen there is a lot of behind the scenes work required. In an ideal world data arrives automatically, is on time and is perfectly structured and there are never any hiccups. In the real world data is messy and doesn't just automatically arrive. Cleaning and bringing structure to messy data is probably the single biggest task any data scientist does. To ensure the data flows as required I'm quite heavily involved with the wider Data Services team in creating the data 'plumbing' to make it all work and ensure the business and customer gets the results.
Interaction with the wider business involves identifying and explaining opportunities and promoting the wider use of data and algorithmic solutions.
Neil Osipuk, senior data scientist at Nine Entertainment Co:
As Senior Data Scientist at Mi9 I get asked this question quite a bit, twice just last week. It’s the first time I’ve had the title in my career, but I’ve been practising various aspects of the job for 25 years across the media, technology and research sectors.
Data science is a multi-disciplinary field, so it’s common to enter from either a software programming, research/analytics or statistics background and then develop the other essential skills. Formal degree programs in data science have only been introduced in the past few years.
I was an industry analyst, strategist, data modeller and market intelligence consultant in Boston, New York and San Francisco before moving to Sydney in 2014. My degrees include a bachelor of science, a graduate diploma in international studies and an MBA in high-tech innovation.
I knew I wanted to move into data science in 2011 after experimenting with emerging data visualisation tools. Using cutting-edge data visualisation tools is now a routine part of my work at Mi9.
My role at Mi9 also involves defining the “single source of truth” for internal KPIs, building reporting systems, testing hypotheses and developing predictive models. Our team answers strategic operational questions from Ninemsn editorial, 9Jumpin or the commercial teams.
Working with Channel Nine’s Research and Business Intelligence teams, we introduced interactive data visualisation dashboards to report on TV campaigns for sponsors of major TV shows, like The Voice. These innovative dashboards show every aspect of the client’s advertising activation by episode, by market, to the very minute.
Richard Hunt, data scientist and senior analyst at Vodafone:
Unlike many data scientists, I don’t have a direct responsibility or reporting relationship to a Marketing function. Instead, our CFO has provided a roving ‘Star Trek’ style manifesto: to explore strange new worlds and bravely go where others fear to tread! We are asked to ‘challenge’ the organisation and the way it thinks.
My job entails a very wide variety of work, often starting with a new idea of how we can add value to the business or improve our customers’ experience with us. For instance our Retail Store network was recently upgraded with digital advertising displays, so an idea around A/B testing different content in different stores was born. It can often take a while to move from concept to fulfilment and financial gains, due to both technical and organisational challenges, but this one is starting to gain legs.
Other jobs can involve looking at customer complaints, modelling these as necessary, or financial modelling for new product releases. The wide variety of mathematical techniques (not just statistical ones) required for my role, and the necessity of keeping up to date, mean I attend industry talks regularly to gain new knowledge to apply to my work. In recent years I’ve been exploring Graph Theory and also Text Mining, among many other things. Another way I keep in the loop is by participating in “Kaggle” competitions (‘statistics’ competitions, often with a monetary prize).
My background includes a Maths/Stats post-graduate research degree, and I think this would be a minimum for anyone starting in this area.
Adam Edgley, data scientist at Optus:
I’ve been a data scientist at Optus for 11 months, but I’ve been writing code since I was in primary school. Tinkering with computers has always been my passion, while my Economics degree has given me the statistical training to turn my passion into a successful career.
So, what does a Data Scientist actually do? We navigate through large volumes of data using a variety of techniques to identify patterns or trends. Put simply, it means taking a different approach to business problems and drawing on data to recognise existing opportunities or predict future outcomes. Data science is about asking why certain data points – called outliers – are different, or whether these data anomalies are a sign of a new developing trend.
Specifically, I look for trends that can help Optus improve the products and services it offers customers. For example, my work can help the business make informed decisions relating to investments in its mobile network by detecting where customers experience call drop outs, recognising when and where network congestion occurs, or identifying when customers’ network demand peaks in a particular suburb.
Data science is becoming more common across a range of industries like banking, insurance, supermarkets. With so many smart tools available to analyse data, our unique skillset lies in how to approach a problem, knowing what questions to ask and how to be creative about solving complex issues.
Data scientists tend to be widely read and have experience with all facets of machine learning, particularly predictive analysis and pattern recognition. Training in computer science, a creative flair and interest in mathematics are useful, but the best attribute is intellectual curiosity.
Jack O'Mahony, manager, information architecture at Red Planet:
Data science’ applies a range of analytic techniques to all kinds of data. Typically these techniques include segmentation, predictive modelling, data visualisation but also possibly traditional statistics and very untraditional things such as machine learning, optimisation and text mining.
Data science is often associated with just machine learning and big data, however this does not necessarily have to be the case and lots of good data science is also done on 'open' data sources, which are becoming increasingly available. Data science can really cover any field from astronomy, to social science, to business, to medicine: as long as you are applying some 'science' to the data involved.
In truth, no single person can cover all of the things that are now included in data science, so the type of data scientist required is very dependent on scope of the investigations required.
At Red Planet we have huge amounts of data which we use to provide better outcomes for our customers. This data varies from mature and established CRM systems through to brand new unstructured social listening tools, through to custom research. It’s my job to take these disparate data sources and integrate and explore them to come up with brand new pieces of information. In other words I’m coming up with answers, to which the questions may still be unknown.
The complexity of the task has increased significantly over the past few years, with data becoming less structured and the volume increasing exponentially. This has provided challenges but also great opportunities - the possible value of outcomes are greater but much more difficult to dig up.
I interrogate data sources where linkages are no longer black and white, but probabilistic and grey. In years gone by these would have been deemed unsuitable, however with the increasing volume of touchpoints with our customers, the sum of these parts is now valuable insights which would previously have been hidden. The insights generated provide real actionable value to our customers through both our advertising operations and professional services arms, while the feedback loops continue to grow the underlying pool of data from which these are drawn.
Stuart Dennon, data scientist at Slingshot:
It’s been 26 years since I completed a research degree in psychology and 24 years since I completed an MBA.
Since then I’ve spent roughly 6,000 days honing my data and science skills in marketing and academia. That’s about 60,000 hours applying the scientific method, experimental design, economics, finance, marketing, statistics, statistics, and more statistics to bring data to bear upon questions important to clients.
I’m delighted marketers have discovered the majesty and power of science and data.
For me, the focus on science is what distinguishes a data scientist from a data analyst or software engineer. Everything I do is calculated to explain systematic variation in the things that matter to clients (e.g. sales) in terms of systematic variation in the things that matter to agencies (e.g. advertising).
At Slingshot, the goal is to find relationships and insights to exploit through media. That means living and breathing client data. I’m a hands on, 10 hours a day SAS programmer and have been for 20 years. I code to extract, transform, and load data into a format that facilitates statistical analysis.
About 15% of my time is devoted to imagining how to use data to ask questions. About 60% of my time is spent manipulating data into the format required to build a model or perform an analysis. About 10% of my time is spent publishing results. The remaining 15% is spent influencing, lobbying, explaining, pitching and otherwise selling my services to agency colleagues and clients. I look to automate processes on the second or third pass, to improve my own productivity and free up my own time. I spend that time off-piste, out of bounds, working beyond the brief on the stuff that nobody has thought of or thinks is important.
If I could say just one thing about data science it would be this. Keep it simple stupid. Most of the untapped value is hiding just below the surface, waiting to be discovered by those who make the time to look.
Have something to say on this? Share your views in the comments section below. Or if you have a news story or tip-off, drop us a line at firstname.lastname@example.org