Making sure data science works for your business

Any corporation seeking to embark on a data science journey needs to consider that a total transformation of culture, people and processes is required.
Read time 4min 30sec

Ever since 2012, when the Harvard Business Review famously labelled data science “The Sexiest Career of the 21 Century”, company executives have been trying to figure out exactly how this revolutionary tool can increase operating efficiency, improve decision-making in the business and, ultimately, boost the bottom line.

To their detriment, many companies have followed the hype, set up data science departments and hired data scientists without thinking the process through, all the while ignoring the potential pitfalls and perils of the data science journey.

Let it be said clearly, though: data science, used correctly in the appropriate business environment, can make a significant difference to the speed and accuracy of decision-making.

Knowledge is power in business and data is the fuel that creates this power. Typically, 80% of corporate data is unstructured and needs predictive analytic tools to gain insights from it. Data science uses scientific methods, processes, algorithms and systems to extract knowledge from data and, by leveraging this data, enables managers to make evidence-based decisions.

However, experience across the world has shown that any corporation seeking to embark on a data science journey needs to consider the following:

Start properly (or re-start properly if you have gone down the wrong track)

The first step is to identify the specific areas/opportunities/needs in the organisation in terms of where data science can add value.

The executive team needs to generate a set of specific use cases to explore. A framework to evaluate and prioritise each use case needs to be set up and the team then lists the target areas to be investigated.

Knowledge is power in business and data is the fuel that creates this power.

Usually, existing internal resources are insufficient to cope with the technicalities, so external resources should be called in if required. All resources should then be allocated to the highest priority situations.

Finally, to deliver value, agile, multi-disciplinary data teams contract with managers through signed-off scorecards and checkpoints to ensure that delivery stays on track.

Understand the process

Essentially, the data science process consists of:

  • Articulating the problem statement clearly.
  • Gathering, cleaning, wrangling and preparing the data.
  • Analysing the data.
  • Visualising results in interactive and intuitive ways.
  • Communicating findings to non-technical stakeholders who need to act on the insights.

Get the foundations done before starting on the building

No business is at the start of this journey – they’re all at varying levels of maturity. Companies often invest in data scientists and, after six months, return with the message that there’s no stable infrastructure on which to do any analytics.

To mitigate this, the organisation needs to have data engineers already in the business, who can work to set up the environment required.

There is no need to agonise over building the long-run big data architecture before the use cases are delivered.

Build the correct team

To deliver value, a team requires a combination of different skills to be effective. Typically, a data team consists of:

  • Analytics translator.
  • Data scientist.
  • Data engineer.
  • Data analyst.
  • Business intelligence developer.
  • Software developer.

The executive team needs to understand the different and key roles played by these various people.

Get the team working at full steam

Like any group of highly talented individuals working together towards a common goal, the team will encounter issues such as leadership, interpersonal relationships, egos, jealousy and ambitions.

In order to get them working together optimally, management needs to pay careful attention to issues of communication, shared goals, motivation and recognition.

They simply cannot be left to get on with the job by themselves.

Where does the team fit in?

Once there is a coherent and competent data team in the business, decisions need to be taken about where it fits into the organisation’s structure.

Ideally, a team should operate as a “centre of excellence” within the organisation, rather than decentralising it across departments or simply placing it in the IT department.

But this will depend on the level of maturity of data science within the company and the particular organisational structure in place.

Finally, understand this is a journey, not a destination

The journey to a mature state of excellence in terms of data management in a company is essentially never-ending.

Technologies change all the time; the organisation’s needs change; people change.

The road to maturity is difficult and existing organisational habits and reflexes need to be discarded to get there. Nothing short of a total transformation of the culture, people and processes is required.

This, of course, is not only the preserve of the data science department. In an era of rapid environmental change in everything from macro-economics and international trade agreements, to local market dynamics and consumer demands, companies need to be vigilant and agile.

And, perhaps, the introduction of a state-of-the-art data science capability will allow the whole organisation to be more competitive, efficient and effective.

Shaun Dippnall

founder CEO of EXPLORE

Shaun Dippnall is founder CEO of EXPLORE, a portfolio of Digital Academies that deliver digital skills at scale to young South Africans in an inclusive, innovative and creative way.

Within the EXPLORE Academies, the scientists research and develop scalable IP that is applicable to global audiences with a specific focus on financial services and utilities.

Dippnall is a qualified actuary and has lectured in actuarial science at the University of KwaZulu-Natal. He has held the positions of chief data scientist for Telkom and was chief actuary for Vodacom and Nedbank. Prior to these roles, he was a business analyst for McKinsey.

See also