Unser Blog

The importance of data science in innovative and data-driven organisations


18 . Mai  2017

Vidhya Karthikeyan

Posts nach Autoren: Vidhya Karthikeyan, Principal Research Scientist, BT

Data science is increasingly becoming a core capability. After all, who or what is going to turn the increasing mass of data into something usable in every department, from marketing to product development? But there’s a real concern because of the shortage of data scientists and the low skill adoption rate compared to the growth in technology.

Our principal researcher Vidhya Karthikeyan shares her view on the importance of data scientists in organisations who want to be data-driven and future-orientated.

Firstly, what exactly is a data scientist?

Data scientists enable evidence-based decision making in companies like BT. This helps to spot unintuitive trends we may not know to look for in the first place.

Everything is changing, constantly – from network infrastructure to the services we deliver over that infrastructure to the business processes that support both. People of all levels in a business have to answer different questions as their company evolves, but data scientists with a wider purview, such as those in research, are able to take a step back from the immediate question, identify common threads across a multitude of similar questions and call upon the right tools for the right question. There is no one size fits all.

There’s more to data science than simply finding patterns in existing data. It’s as much about tools and artificial intelligence (AI) as it is about domain knowledge, trust in stakeholders (and vice versa) and an understanding of what makes actionable insight in the business at that time vs. interesting but not necessarily actionable information.

Data scientists help the company define the way they collect and treat data to ensure that those efforts lend themselves to innovative solutions and insights for the business. Research data scientists help companies develop a tools roadmap. As pioneers of innovative proof of concepts, they are exposed to a number of tools that fit different purposes. As organisations evolve as a whole, a data scientist is well-placed to inform decisions on tooling for the wider community with experienced views on ease of use and capability with a potential target audience or infrastructure in mind.

Downstreaming data science is unique and differs from traditional software delivery. Data scientists might derive insights from data and present those insights to drive better decision making within the business. They might design futuristic learning algorithms that are then integrated into existing IT systems to enable automated and proactive action to improve customer experience. They might assemble a capability from a number of existing tools, using slivers of each of those tools and transforming data from one environment to another. Right from wrangling data to modelling to applying machine learning for discovery, the output and skillset of a data scientist can vary vastly to reflect the needs of the business.

And what does good data science look like?

In my opinion, good data science should be three things:

1. It must enable better and faster decision-making. The main purpose of data science is leveraging the vast volumes of data we have always collected and transforming an existing capability into one which is data-driven. Often this means changing existing processes as a result of the insight you’ve got from the data.

2. It should think about the common capability that would address a number of related questions rather than design spot solutions to analytical problems.

3. It must also be reproducible by others. Firstly this is to ensure quality control. Secondly, whilst a data scientist might design a data solution and help integrate it into the business, the maintenance of those solutions might reside in another part of the company entirely. Good data science is documented well enough for it to be understood by someone in the organisation with a different skillset entirely.

Aside from the benefits of capturing actionable insight, what other benefits can companies see where they have strong data science capabilities?

There are a number of other benefits. For example, data scientists can bring an enterprise closer to its customers - we would be better in tune with our customers if we designed our systems with data collection and data architecture as a primary focus. That’s why having a common data lake where data is easily accessible by everyone (within regulatory boundaries, of course), from marketers to finance teams to product teams, with appropriate toolsets that are accessible to each of those audiences, is vital to ensuring success.

Data scientists are often asked to look at a variety of application areas based on common data sources and people like myself are involved in steering the business towards better data architecture, better reusability across applications and better value from data overall.

What advice would you give to organisations who want to become more data driven?

Fundamentally, organisations should spend substantial effort when designing new products, services and business process to decide how their data is collected, handled and retained. Data that runs a daily business-critical system is different to that which powers analytics.

Allow your data scientists to search and fail. Create IT infrastructure that supports daily business-critical applications as well as those heavy duty analytic applications. A trend that isn’t visible when zooming into today might be visible when stepping back in time. Document your data and your data science – it may not the sexy part of the job but certainly a necessary one. Trust the data: organisations have migrated from manual to automated – why not to autonomous?

Can you share an example of where data science has truly impacted another part of BT?

BT collects volumes of data about how well our networks and services perform. Operation teams are inundated, and extracting information from data is no longer a manual process. I’m quite proud of an early proof of concept activity I did, where I designed an algorithm that predicted with a high degree of certainty the likelihood of a failure based on device temperature-related events. These precursor events were previously being ignored by the operations teams simply because they was no way to know that there could be a relation till my work discovered the link.

One of my areas of research is to use service performance and network data to better deliver TV to our customers. By applying data science and machine learning techniques, my work has helped diagnose various root causes of TV degradation in the past across our nationwide infrastructure. It has also enabled more informed conversations with our customers when they get in touch with us. Some of my work recently underwrote our TV growth strategy to increase the number of channels delivered over our network by ensuring that we have sufficient network capacity to carry that traffic. My research enables our customers to watch high definition TV in good quality their living rooms and on the move.

Finding data scientists can be tricky for organisations. What tips would you give in building up a strong data science team?

Recruiting a good data scientist requires knowing how to spot one when you see one. I wouldn’t claim to be able to evaluate a candidate until I have given them a data science task with some data and listened to them think through how they might approach it, if not actually do it. Tools can be learnt sooner than a way of thinking. Data scientists learn the ability to communicate their technical findings through experience – it isn’t easy to turn a mathematical model into words that everyone understands.

Organisations should encourage their young data scientists to have that exposure to the rest of the business as it helps them identify a common problem space and communicate their ways of addressing it. Allow your young data scientists the space to play with data and tools in a sandbox environment without necessarily asking a direct question. Encourage creativity and help find avenues to guide their findings to deliver business impact as much as directing them to answer questions that interest you.

Vidhya will be taking to the Innovation 2017 stage to talk in more detail about the role data science plays within modern organisations.