Data science has become an inescapably critical facet of business operations across practically every industry in our data-focused age.

The role of the data scientist within a business is to use AI and machine learning to convert raw data into deeper insights that can help to benefit an organisation.

The potential benefits of good data science practice to your data management are numerous, as well as being useful for uncovering actionable insights within your data, which can then be used to guide decision making and strategic planning.

When considering what positive impact looks like, we must make it measurable. Business owners must establish what these positive outcomes could be, and this must become a continuous process to enact positive change.

Leanne Fitzpatrick, director of data science at the Financial Times, took to the stage at this year’s Digital Transformation Summit to provide a deeper dive into some realities of making an impact with data science, and ways leaders could strengthen the process within their business.

“When we think about our data, we need to have good foundations. That means enabling the data science team to be able to select the right data, source it, and synthesize that information,” she commented.

Fitzpatrick viewed three areas that must be “coupled together” to improve the process of data science, and leaders should focus on technical implementation, the data science team culture, and an organisational embedding of required outputs.

Technical implementation

First, Fitzpatrick discussed the implementation of such a technology, and whether it is currently possible for you or your business to implement it effectively.

“What I’m talking about here is the realities of the capability to deploy and develop data science, machine learning, and put that into production,” she said.

“That can be your data, your model, and your code management. It can be open-source toolkits and languages like R and Python and being enabled to put those into production.”

She noted the important of being able to scale and load balance: “We’re in the era now of mass GPU and looking at image processing, natural language capabilities with machine learning.

“Those take a huge amount of memory resource, being able to scale and load balance that so that in real time or in production you can get a response version control and lineage,” she said.

Data science team culture

Fitzpatrick moved on to focus on building a data science team culture, which puts much more of a focus on humans and people.

“What we are looking for in our people is trust, curiosity, and knowledge. Why will you need the knowledge to be able to find the right data and business problems to solve? You need to be curious to go out and find those,” she said.

“You need to enable a culture of trust where people can identify business challenges, identify business problems to be solved with data science, and enable those people to get on with it.”

Additionally, she noted that firms must consider other areas including building a good model and code review processes, strong documentation, collaborative teams, and squads so that things done deprecate, and that things don’t grow in silos.

“Mentoring and coaching, cross functional upskilling, and the correct people in skills,” she said.

Organisational embedding

In terms of organisational embedding, which Fitzpatrick noted is one of the most difficult to implement, she put a focus on changes in decision-making behaviour.

Part of data science process will be the outputting of information, and she suggested that there should be a decision-making process off the back of it.

“We need to have good governance and compliance around how the organisation is using these prescriptive and predictive models,” she continued.

Recommended

• BBC reports £150k of electronic devices stolen over three years
• Embedding innovation into all aspects of business strategy and culture
• Introducing the inaugural Scottish Black Talent Summit

“There are still a lot of sensitivities around our data, and particularly around personal data, the org structure in terms of our organization and understanding how data science can make a difference to different parts of the business.

A focus must also be put towards “competing priorities,” Fitzpatrick said, which involves ensuring that it stacks up to your business strategy and value recognition and attribution.

Fitzpatrick concluded: “Making an impact in data science isn’t just about this technical implementation and what we’ve coined that as ML Ops, we need to bring all these things together.

“That’s what I refer to as machine learning operations is when you bring the cultural and the embedding together. That’s the only way we are going to continue to make impact in our businesses.

Get the latest news from DIGIT direct to your inbox

Our newsletter covers the latest technology and IT news from Scotland and beyond, as well as in-depth features and exclusive interviews with leading figures and rising stars.

To subscribe, click here.