Governing Data

Article written by Pete Crawford

Moving from Principles to Practice.

Let’s face it, data governance has a reputation of being a worthy, essential, but staid topic – a necessary prop underpinning aggressive innovative strategies or new analytic frontiers such as automated decision platforms powered by Deep Learning models. Part of this perception stems from a traditional notion that the function of data governance is to maintain data quality and reduce risk by upholding data protection regulations. By adhering to this formulation, governance is primarily viewed as a set of ‘command-and-control’ rules with escalation points.

A more balanced view is to regard data governance within the wider perspective of value creation. With this lens, governance becomes an important extension of developing data literacy skills and task-based, ethical accountability throughout the organisation.

“Insights from Pete Crawford | Head of Data, Analytics & AI, Pete Crawford spends his day-to-day leading strategy, governance and execution over enterprise data platforms, data science and AI capabilities. Speaking at leading industry AI and data events, Pete is experienced in forming and directing multi-disciplined teams to manage enterprise information assets and deliver business outcomes through advanced analytics.”

The journey from principles to practice starts by outlining why governance is central to any data strategy or business transformation endeavour. An assessment of what sort of governance model is most applicable to the power dynamics of the organisation follows. This leads to key milestones for establishing a governance program and the types of activities that support a well-informed and data privacy compliant playbook.

Why Data Governance Matters

Strategic drivers

• Changing regulatory requirements with data protection and privacy – which carries large financial penalties for compliance failure.

• Customer experience – the outcome from improved operational decision-making stemming from superior interpretation, sharing and evaluation of data.

• Accountability – linking the quality and trust of data with an assurance that employees responsible for business outcomes are part of its governance.

• Reputation management – by containing potential data breaches which invariably lead to costly PR fallout, weakened business valuation, customer churn and consumer abandonment.

Tactical drivers

• Data is available, discoverable, consistent and appropriate to different domains of users.

• Guidelines for the acquisition, sharing, processing, combining, retention and deletion of data.

• Capture of all consumer data event journeys coupled with consistency of interpretation across all data consumers.

• Prevention of bias and discrimination which may be inherent or inadvertently introduced through data, algorithms or practice.

Governance Structure Options

Item	Top-Down	Bottom-Up
Summary	~ Leads by authority ~ Hierarchical	~ Democratic control and management of conditions ~ Shared vision and principles based on network relationships
Principles	~ Standardised rules ~ Dedicated data stewards	~ Everyone has clearly defined accountabilities
Operating Model	~ Defensive - risk mitigation focused	~ Agile – policies and processes closer to the end user
Suitability	~ Rules are set by executive branch or dominant business unit who owns the business problem and proposed solution	~ Data is treated as a shared strategic asset across multiple business functions
Risk	~ Lack of context or sense of ownership over the application of rules	~ Coordination complexities across multiple business units

Setting Up a Data Governance Program

A governance program – like any change initiative – should begin by recognising the importance of storytelling. More specifically, there is a need to rethink and reframe narratives around enterprise use of data that look beyond ownership, protection or unquestionable economic value in order to bring into focus concepts that validate access, trust, agency and partnerships. Key considerations include:  

Awareness

• Ensuring executive sign-off and visible sponsorship of vision, principles, structure, funding and timeline.

• Understand, up-front, key messages from a strategic, tactical and operational perspective and how they should be differentiated according to role or domain.

• Be clear with stakeholders that there are time, resource and budget implications – too often governance is assumed to be a component of BAU, or worse, achievable by simply standing-up a committee.

Roles and responsibilities

• Establish and clearly communicate who is running the program. A common challenge is that people assume IT ‘owns’ the data.

• Identify data domains.

• Form a governance council from business leaders and partners.

• Identify operational data stewards. Stewardship is a trained and formalised accountability which describes a task-based relationship to data. It is not a hired position – anyone can be a data steward.

Standards, policies and processes

• Commence discovery to identify critical pain points for what business units cannot do because of a lack of availability, quality or knowledge about data.

• Review and consolidate existing policies and practices that define enterprise data engagement behaviours.

• Define missing policies e.g. how are algorithms being monitored for fairness?
  
  

Value creation

• Take a human design-centric approach by engaging with data consumers, both inside and outside the organisation, to recognise their aspirations and pain points when dealing with their ability to share, use or retrieve information.

• Educate stakeholders by translating governance principles into business context.

• Formalise data literacy programs by focusing on improving how employees:

  • Use numbers, statistics and infographics to convey important messages;

  • Evaluate data collection or automated decisions for bias and discrimination;

  • Use data analytics, find insights, identify trends and make decisions;

  • Use data platforms in a self-service capacity;

  • Link and share data without compromising privacy or proprietary.

• Clear associations are established between data quality, data usage and customer experience. Measurable incentives should form part of a group’s performance evaluation.

Barriers to success

• Lack of leadership – or, conversely, total reliance on top-down command. Leadership needs to visibly support the program and reward team accountability over data.

• Lack of investment – to counter, a simple cost-benefit exercise can help set a baseline against the costs of compliance if data governance is not implemented.

• Business units retain a proprietary sense of ownership of data – the breakdown of silos needs to be central to a coordinated data strategy and modernised data architecture plan.

Creating a Governance Playbook

A clearly defined playbook is required to put a data governance program into action. Each activity should be clearly documented, communicated, frequently updated and referenced to relevant regulatory or ethical standards. Some selective activities are listed below.

A playbook can be exhaustive, but if starting from scratch then concentrate on:

• Start small by focusing on a particular business unit or data domain.

• Define business ownership and identify roles and responsibilities.

• Map data flows across infrastructure and to operational tools.

• Place data education and task-based accountability at the centre of the program.

• Set measurable goals (especially around end user experience).

Strategic activities	Operational activities
~ Vendor risk assessments ~ Data sharing agreements ~ Data broker or marketplace evaluations ~ Establishing decision rights ~ Issue resolution and approval path ~ Acceptable use and consent standards ~ Technology platform options ~ Communications plan ~ Measuring and reporting value ~ Data collection bias ~ Algorithmic fairness	~ Data taxonomy and classification ~ Data collection standards ~ Data quality specifications ~ Data lineage and data flow maps ~ Data masking standards ~ Data privacy impact assessments ~ Issues register and matrix ~ Privacy-by-design ~ Differential privacy ~ Model registry and feature stores ~ Automated decision observability

Evaluating Governance for Short-Term Effectiveness and Long-Term Value

Business Impact Metrics Examples	Operational Metrics Examples
~ Compliance cost ~ Application development cost ~ Customer satisfaction ~ Data quality	~ Data governance maturity level ~ Data management efficiencies ~ Data literacy ~ Data governance issues register

LOOKING TO Leverage and utilise your data? REACH OUT.

Whiteark is not your average consulting firm, we have first-hand experience in delivering transformation programs for private equity and other organisations with a focus on people just as much as financial outcomes.

We understand that execution is the hardest part, and so we roll our sleeves up and work with you to ensure we can deliver the required outcomes for the business. Our co-founders have a combined experience of over 50 years’ working as Executives in organisations delivering outcomes for shareholders. Reach out for a no obligation conversation on how we can help you.

Article by Pete Crawford