Around the world, companies are grappling with an avalanche of data from a growing number of sources. Yet it is not the volume of data that matters, but one’s ability to rapidly convert data into actionable insights and build accurate predictive capabilities.
At a panel discussion “How Data and Cloud are Changing Business Fundamentals” at the 3rd CDOTrends Digi-Live! Summit held in January, panelists discussed how organizations can get their data initiatives started quickly, and offered tips on getting the right data science talents.
How to start quickly
So how can organizations start deriving value from data? Existing data and off-the-shelf data tools might be a good place to start, according to Juliana Chua, the head of regional digital transformation at the Essilor Group. A self-professed fan of data experimentation, she pointed out that there are many ways that organizations can derive value from their data without the costs and overheads of setting up a data warehouse.
“Before we look at [setting up a data warehouse for analytics], there are tons of Excel spreadsheets that everybody is [already] collecting. And there are quick ways that give you [insights]. There’s Google Data Studio; there are these free features that you can use to gain detailed insights.”
It is vital to ensure that the data is simple to understand, she notes. “The trend tells us a story, and the story enables the team to make decisions about the correct direction to take. Another strategy for a quick win is to leverage market insights. Beyond data, we also look to our people on the ground to see whether the data matches up with reality.”
While harnessing value from data looks easy, getting it done correctly can be tricky, observed Jacqueline Teo, the chief digital officer of HGC Global Communications, who also suggested that organizations begin with existing data.
“Start with what’s already measured. Your company has more than enough [metrics] that they measure, [but which] they don’t have sufficient analytics for. It is about adding value and trying to move the dials up with incremental value steps. E-commerce sites, websites – there is more than enough analytics there,” she said.
Finding the right people
When it comes to making sure they have the right talents and skillsets, should organizations train a business professional in data science, or hire a data scientist? Panelists have differing ideas on this front, reflecting their diverse experiences.
“To train business professionals who don’t have the computer engineering or analytics background to be a data scientist is challenging,” noted Garrett Teoh, who is also an analytics instructor at General Assembly and adjunct lecturer at the Singapore Management University.
“They need to understand the math behind the data science; they need to understand statistics. They will also need to start writing SQL queries to extract the data, not to mention the more complex data science algorithms, and then pull the output into visualization platforms such as Tableau.”
In Teoh’s opinion, a data scientist from a similar industry who is already aware of the industry jargons and technical lingo, but with no experience in the organization might be a better long-term option if they are brought up to speed quickly, than attempting to train business professionals into data scientists.
And while having the right tools can help, it might still take too long for them to sort through the data science jargon, much less dive deeper into topics such as machine learning. “Without understanding the foundation such as algorithms, [data science] is difficult for them to decipher,” explained Teoh.
Training employees… and its limits
Sourabh Chitrachar, the regional director of IT transformation & strategy for Asia at Liberty Mutual pointed out that while the preference is to hire someone with both the domain knowledge and the data science background, this is challenging to pull off in practice. On this front, his organization had hired people externally, as well as identifying and training “power user” employees with some familiarity with data.
“The person who is trained already has a thorough understanding of the domain. With the right tools and knowledge [being imparted], they can come to the right level of expertise in due course. It takes time, but it has worked,” said Chitrachar, who also cautioned that this strategy won’t work if results are required quickly.
For Teo, the challenge isn’t so much about hiring competent data scientists but ensuring that they have the acumen to thrive in a particular industry. “Finding the right data scientists can be quite scientific, and you'll normally get enough candidates with enough skill sets. What we find very hard to train, and which people struggle with, is getting someone to recognize an opportunity.”
“You can teach someone how to validate the algorithms, but it's hard to teach someone how they need to work within the business processes, and to have the business appetite, in an organization,” she said.
Roadmap to success
Chitrachar offered two steps for success, namely having a clearly defined scope, and being focused on the outcomes of data initiatives. He said: “Without a proper strategy, it’s just not possible to succeed. Unless you narrow down the scope and be very clear about what you want, and what value you get out of that data, it becomes impossible to navigate that whole ocean of data.
“It is vital not to get carried away by the desire to boil the ocean by doing everything out there. We do PoCs [Proof of Concept] to better understand the value of certain projects. We then go back to the CEO and senior executives to verify if the benefits were what they expected before proceeding further.”
“If necessary, we tweak our projects following an agile methodology. You fail, fail faster, and move quicker. This approach has helped us to formulate and fine-tune our data strategy and execution.”
Watch the full panel discussion here.
Image credit: iStockphoto/metamorworks