Becoming Data Centric

I’ve spent the last two decades working with analysts to solve data problems in a systematic way and to create platforms that give analysts easy access to the data they need. I like to think that I’ve gained a good understanding of the kinds of problem that analysts typically encounter, of how those problems arise and persist, and of what a good solution looks like.

You don’t solve these problems with technology alone - you solve them with technology, people and processes working together in a combined data centric approach.

The Application Approach

It’s difficult to articulate what a data centric approach is without first describing the 'Application-Centric' approach that has been prevalent for decades and remains the status quo in many Enterprises.

An application-centric situation is simply what arises organically over time when an IT landscape evolves through the sequential implementation of many individual projects with different business goals. Projects are typically sponsored by the heads of a particular business function and are focused on fulfilling project requirements primarily for their own interests, often with limited regard to broader Enterprise needs.

Broader, strategic concerns for the enterprise will only come into play if they are imposed on the project from above or by external forces, like regulation.

If nobody is enforcing an alternative approach, this will over time naturally result in a proliferation of individual 'point' solutions, implemented using a variety of technologies and each hoarding its own data like a miser hoarding pennies. It is this situation that gives rise to the problems that data analysts are all familiar with.

Impacts for Enterprise Data

Most organisations have many applications, so the immediate effect is a decentralisation of data from the core to the periphery, with duplication, redundancy, and inconsistency of data, and with an inability to connect data silos across the enterprise.

An application typically embodies a narrow view of data, reflective of its own specific needs. It stores data in an idiosyncratic manner without much regard to how other processes or external functions might access or use that data in future, meaning that application data is often difficult to get at, and difficult to make sense of outside of the application. Another consequence is that data becomes discordant between applications, with different codes, formats and meanings in different systems.

Applications all communicate using the same medium – data, but they often don’t understand one another’s use of it.?

The Data Centric Approach

Being data centric means considering the data first, and then building the application landscape around it. At the core are four fundamental concepts:

1. Data is a key strategic enterprise asset, that is fundamental to good decision making, that generates value and is of value in its own right.
2. Data is the critical input to and output from applications. Applications should adapt to cater for the data, not the other way around.
3. Data has a long lifespan that often predates and routinely survives the lifespan of individual applications.
4. The use of data will extend beyond immediate business needs and will be used at an enterprise level in ways that the original application designers had no inkling of.

Accepting these 4 key tenets will lead rational decision makers to take an approach that ensures enterprise data requirements and needs are given a central consideration and priority in all plans.

Key features of a data centric approach are:

1. Centralised enterprise data models, that are not dependent on any specific application, and will remain in place and valid as the application landscape evolves and changes around it. This core set of coordinated data models will share meaning and draw together the different elements of the business in a consistent and clear way.
2. A framework of data governance and management policies and standards that define how data will be generated, shared and controlled.
3. An organisation structure with resources in place to monitor and enforce the application of the agreed standards, and empowered to take corrective action where necessary.

Benefits of a Data Centric Architecture

A data centric architecture delivers a simplified data landscape and data flow map with no inaccessible, cryptic or isolated silos of data. It is flexible, scalable and powerful, with 'master' data consistent across the enterprise, and with data quality managed and improved by a central governing team.

The consistency and synergy delivered by this approach empowers enhanced reporting and analytic capabilities for the enterprise and reduces costs and risks associated with data. At the same time the unified governance of data and processes provides clarity of purpose and execution across the enterprise.

What Does it Involve?

Becoming data-centric requires:

• Dedicating resources to planning, implementing and maintaining the data models
• Revising project lifecycles and developing a system of rules and policies to ensure that application data is integrated to the central core model as a matter of course, not as a secondary consideration.
• Implementing monitoring and controls to ensure the data exhibits high quality in all dimensions -accuracy, completeness, integrity, timeliness etc.
• Getting a firm grip on existing master data across different applications, departments and regions and having procedures in place to ensure new sources of master data can be seamlessly integrated into the whole
• Documenting data models and processes with sufficient metadata and supporting information that they can be easily understood and used by the business – model diagrams, data catalogues, business glossaries, data flow maps, data lineage, semantic layers
• Establishing appropriate governance structures and oversight practices to ensure policies are maintained and adhered to.

Data entropy is pushing constantly for the application centric outcome, so achieving a data centric architecture requires on-going effort to wrestle data back from the edges into the centre.

Organisational discipline is important. If a new application cannot provide its data in the required manner, then its implementation should be challenged.

Manual processes are subject to variation and error, so where appropriate, it can be useful to invest in tools and software that reduce the manual element of data management tasks. A wide range of tools is available that aids in the control of data quality, the maintenance of metadata and the management of master data.

Conclusion

A change in emphasis like this involves a change in culture – and as the famous adage goes, 'culture eats strategy for breakfast'.

Data initiatives require proactive collaboration between IT teams and business teams - so foster that collaboration. Work to persuade sceptics before imposing new rules upon them. A good way of doing this is to establish a data strategy that has involved and included all the key stakeholders. The data strategy should include a roadmap outlining key programme components and projects so that these can proceed subsequently without dispute or controversy.

Executive mandate is crucial. The executive’s job is to develop a long-term vision and make strategic decisions for the enterprise that go beyond day-to-day needs. Individual business units are rightly focused on their own local goals with little interest in enterprise initiatives that won’t directly benefit them. Hence executives must sponsor initiatives from the top down, make the necessary decisions and make available the necessary budget to prioritise data as a strategic asset. Again, a formal data strategy is good way of securing this from the off.

I hope readers will see the benefits of the data centric approach. It covers a broad set of initiatives and activities, but even doing some of them will be beneficial.

Questions on Data Warehousing, Data Integration, Data Quality, Business Intelligence, Data Management or Data Governance??Click Here?to begin a conversation.

John Thompson is a Director with EY's Technology Consulting practice. His primary focus for many years has been the effective design, management and optimal utilisation of large analytic data systems.