Anatomy of data governance

The proliferation of data is driving enterprise-wide digital transformation initiatives and business decision-makers are aligned to the urgency of leveraging their data for sharper insights. But how can they effectively dissect their data and unleash its true potential?

Data Governance (DG) is a central theme for most organizations, as they battle with the core challenge of instilling confidence in data among the business user and the ability to access and sustain effective and comprehensive data assets. Good data governance will allow organizations to drive outcome-based data science programs along with high-quality descriptive analytics.

Challenges of the past such as high degree of manual intervention, inability to identify levers for causal analysis, multiple versions of truths, limited technology maturity, and an ever-evolving application landscape among other factors precluded organizations from creating a reliable data management platform. Eventually, organizations slip into data chaos and business users make decisions based on gut instincts, moving away from being analytics driven.

The prospect of an unscientific approach is terrifying, and every attempt should be made to eliminate all forms of bias in decision-making (confirmation, anchoring, retrievability, regression fallacy, hindsight, hyperbolic, among others) by truly embracing a data-driven approach across transactional and master data.

A DG council supported by data owners, data stewards, and data custodians is critical in making analytics programs successful. Key business and data stakeholders that have accepted the responsibility must feel the consequences, if the systems and processes they manage are unable to measure or process large volumes of ingested data across the organization. Without a performance cudgel, these individuals may become one of the weakest links in the chain. It would, thus, behoove every organization to appraise these individuals against key progress metrics that also provide a measure of how the enterprise is performing on overall governance:

• Business KPIs (say, OTIF for Inventory Management) that are scored for their accuracy and ability to identify quality gaps 
• DG KPIs (such as coverage of masters) that enable DG council members to take decisions regarding their progress
• KPIs to ensure consistency across data applications maintaining transactional integrity (such as pricing information distributed across applications)

At an enterprise-level, data management needs to address the following:

• Increased pace of data ingestion (structured, semi and unstructured data generated from both within and outside the organization including third party data)
• New data sources (organic or acquisitions)
• Evolving data regulations domestic and international
• Differing priorities for common data assets across functions (impact on pace of change)
• Culture (self-service, data driven, business outcomes, data COE’, human elements)

TheAforementioned aspects have been the primary focus for most organizations, which have been inundated with data (IOT, Social, Edge). Interestingly, enterprises invest in ingesting data, though there is limited progress on governing. Primary challenge is the limited appreciation on how to treat data by decision-makers in the CXO suite. Data governance is two degrees removed from arriving at ROI.  This complicates the decisions to invest in a comprehensive data governance program.

Most data governance programs would be deemed successful if they address: 

• Setting policies and procedures
• Data Quality (identification/ remediation)
• Technical Metadata and Business Glossary
• Attribute standardization
• Data Lineage
• Operating model
• Implementation (various forms - centralized, federated, hybrid), and 
• Integration (data life cycle - publisher/ subscriber)

While the aforementioned steps are critical to setting up the foundation, they are unable to address the issue from a business users’ perspective. A DG program’s primary focus should be to eliminate data trust issues that arise from the consumption layer. Providing DQ scores at the KPI level, and all hierarchies within a given master (enterprise, reference) is an important step that will inspire confidence within system users. This will help accrue several potential benefits:

• Eliminate data trust deficit by introducing reliable quality index
• Improve issue management resolution process by extending detailed causal analysis
• Integrate backend foundational items to visualization layer that will enhance transparency between consumption (business users) and producers (IT) 
• Enable self-service with greater confidence due to high data quality
• Leverage one version of truth for effective decision-making 
• Avoid any data leakage and establish delicate balance between data creators and subscribers

Ant societies are a way to relate policies and procedures followed by them. Most are ruled by a single queen who rules the colony while breeding new generations.  She defends her position with the aid of high-ranking enforcers.  Circle of security allows the queen to overpower any breach of understanding by an aspirational rebel.  Ant society has this policy defined and effectively monitors and reacts if deviations are observed.  However, it doesn’t get into the mechanics of how to do it.  Essentially, the purpose of data governance (DG) is to ensure data is secure, private, accurate, available and usable.  Without this as the guiding definition, every program will have blind spots, resulting in limited resonance value for business constituents.  

Data governance has the dubious distinction of being a priority, but with few qualifying to declare success. Organizations should retain atavistic data governance features (securing approvals, linking data to business transformation programs, setting priority, conducting change management programs on the topic) in addition to the introduction of new elements. It implies incremental innovation, facilitating governance through the lens of the business user, supported by custodians of foundational systems. Fortunately, we have now been ushered into an era where whole data population is a reality and should provide the business constituents their due without further ado.

The article has been written by Chandrasekhar Mukkavalli, Partner, Digital Lighthouse, KPMG in India