By Chris Probert, Partner & Head of Data, Capco
As the volume of data produced and stored by organizations increases, advanced analytics such as machine learning have a crucial role to play in turning that data into meaningful business insights. At the same time, organisations face significant challenges due to the rapid expansion of data types. If they are to capitalise successfully on opportunities, firms must overcome the shortcomings inherent in data silos, as well as addressing poor visibility around data journeys and the ‘clutter’ of data entropy.
- Data silos – unification in the age of analyticsData is the core asset for financial services firms, yet most of is heavily siloed across disparate and isolated systems, departments, functions, geographies, databases, files and archives. Even the simplest requests must be processed via multiple channels and business areas, each accessing a variety of data sources.
The key is ensuring data is available at the right time, in the right place, and in the right format. To achieve this, firms must unify their data silos using an analytics platform. Such platforms are fundamental tools for harmonising data, enabling firms to capture, store, compute and analyse data from a variety of sources in the most effective and efficient fashion.
- Data lineage – a clear line of sightIn a world of big data current techniques for assessing information risk and controls fall woefully short, as they do not provide adequate visibility into the data. A tool to track data from point of origin through transformation(s) and ultimately consumption, data lineage can fill this gap. However its potential is untapped due to a lack of industry standards.
There are three critical guiding principles to make data lineage useful and standardized. One, make it business friendly. Two, highlight context and ownership of data. And three, show how data is transformed and used. Firms that successfully leverage data lineage will derive added value through cost reduction alongside risk mitigation via enhanced controls.
- Data entropy – clearing the clutterThe reluctance of organizations to retire or purge data leads to overflowing repositories and storage packed with outdated, unseen and difficult to access information. In other words, data clutter. Temporary fixes only add further layers of clutter and drive data entropy – the tendency for data within an enterprise to become increasingly disordered.
While large programs are often data-centric and fuel data clutter, they also present an opportunity to implement market-leading practices around data management. An organization that instils a culture where data is viewed as an asset will be able to ensure that every large program serves to enhance its data ecosystem.
- Machine learning – embracing automationMachine learning (ML) can autonomously identify patterns, analyse data and interpret it via reports and data visualizations. ML can be a powerful addition to any data analytics toolkit – but careful planning and a high-level understanding of the relevant techniques are required.
The business question an organization is seeking to answer will determine the type of ML to be deployed; and the type and quality of data available must also be key considerations. If firms can successfully unify data across silos, introduce data lineage to increase visibility, and use large programs implementations to reduce entropy, then the opportunities presented by ML are significant.