3 key ingredients of a successful data science project
A growing number of companies have embraced big data, and data science now plays a significant role in many organisations. A 2017 study from Continuum Analytics found that 62% of companies use data science on a weekly basis (with 31% using it daily), and that 89% of enterprises have at least one data scientist responsible for applying data science to the business. The upward trend of big data adoption is confirmed in another 2017 study from Dresner Advisory Services, which found that 53% of companies today have adopted big data, up from a mere 17% in 2015. And at AT Internet, nearly two-thirds of our customers have initiated a data science project!
Data science is claiming a critical spot among companies’ priorities and will become an increasingly important lever for business performance. Whether you’re new to data science or already running innovative big data initiatives, learn what the three keys to a successful data science project are. (Be sure to check out the replay of our webinar with data scientist and strategist Dr. Sébastien Foucaud, “Harness the power of data science”!)
1. Relevant and high-quality data
No surprise here: data is a central pillar of a successful data science project.
Obtaining the data isn’t a problem — from e-commerce to media to telecom, no matter the industry you work in, you likely have massive amounts of data at your fingertips: digital analytics data, transactional data, CRM data, offline data, social media data, and the list goes on… There is no shortage of data today, and it continues to be produced at an exponential rate.
But don’t be misled by the “big” in “big data”: simply having access to a hoard of data is not enough. For your data science project to be fruitful, you’ll need high-quality data that is relevant to the problem you’re looking to solve.
“When we talk about big data, the misconception people usually have is that you need a lot of data,” says Dr. Sébastien Foucaud. “But actually, volume isn’t always a necessity. In fact, it’s better to have relevant data that gives information about your users or your transactions, more than just ‘a lot of data.’”
Your digital analytics data can be an excellent source of relevant and high-quality data (especially if you work with an analytics provider who delivers completely reliable, unsampled data). Why? It provides ultra-detailed, exhaustive and consistent information about how users interact with your brand across devices and platforms, down to the individual visitor. And tools like Data Flow make it simple to extract these millions of events, clean and enrich the data, and then feed your data science and machine learning projects in just minutes.
So remember: relevance and quality over quantity. And with your digital analytics data capital, you’re already off to a great start!
→ Learn more about why digital analytics data is a fundamental data source for big data projects.
2. Powerful (yet accessible) technology
You’ve got the right data to fuel your data science project. Now, you need the technology and machine intelligence — the second ingredient — to do the heavy lifting for you.
Let’s take a step back to first answer a simple question: why should we trust a machine to exploit our large data sets?
“Machines are better [than humans] at recognising patterns,” says Dr. Foucaud, citing the stock market or weather forecasts as common examples where “historical data is used to exploit trends and patterns in order to predict the future.” With the upsurge in use of deep learning, this pattern recognition technology is reaching new levels visible in our daily lives. Case in point: highly accurate facial recognition, whether it’s used to unlock a phone, or to identify someone rapidly on CCTV.
Another area where we’re no match for machines is optimising complex problems. “Computers are routinely used today for optimisation — for instance, traffic optimisation, or in logistics to enhance the delivery of goods,” Dr. Foucaud adds. When it comes to meeting specific requirements in the most efficient and/or lowest-cost way, machine intelligence always wins.
OK, computers are great at pattern recognition and optimising. (Good for them.) So why is this technology such a critical element for your analytics and data science projects?
It means we can go beyond “descriptive analytics” (a diagnostic of what’s already occurred) to benefit from “predictive analytics” (an idea of what will occur next) and even “prescriptive analytics” (how to influence what occurs next). Naturally, as the machine detects more effective levers for improving performance, you draw exponentially greater business value from your data.
In the past decade and a half, machine learning and algorithms have become much more accessible, lowering the bar to practicing data science. It’s no longer a question of having access to proprietary technology; a wealth of open-source libraries exists to keep hungry data scientists well fed.
3. The right type of data expert
Having strong-quality data and adequately powerful technology will provide a solid foundation for your data science project, but they are not enough. The third key ingredient, which should not be underestimated, is human intelligence. In other words, your data scientist.
According to the study from Continuum Analytics cited above, 56% of companies find it challenging to hire the right data scientist(s). (Apparently, demand shot through the roof after news of that “sexiest job of the 21stcentury” title got around!) The hype is proving to be true: everyone wants a data scientist, and it is not necessarily easy to find (the right) one.
To make matters even more complex, data scientists can have rather distinct specialisations, which means not every candidate will be a great match for your data science project.
“There are several breeds of data scientists — there’s not just one type,” says Dr. Foucaud. “There are those who can derive insights from large amounts of data and work with business stakeholders… and those who are more machine learning engineers who can encode complex algorithms in your platform and work with your product team,” he says, adding, “Typically, these two types of profiles are very different.”
It can therefore be challenging for companies to determine which type of data scientist profile they truly need (keeping in mind that this can vary depending on the specific data science project in question). One solution is to use a digital expert marketplace like certace to match vetted data science talent to your project needs. Many data scientists prefer to work on a freelance basis, as it allows them to remain flexible and selective about the projects they accept. Using a marketplace to source your data scientists means you can access these freelance experts on demand and ensure the best skill match for your different projects.
Whether your company is just dipping its toes into data science or is already seeing profitable results, you can be sure that big data and machine learning will continue to be major priorities for many companies in the coming years. Learn more about harnessing the power of data science for your business (and some practical use cases!) in this webinar replay:
Originally published at blog.atinternet.com on February 14, 2018.
