Managerial Decision-Making and Digital Transformation

Introduction

Business intelligence (BI) has evolved from periodic management reporting into an organizational capability that connects data, analysis, decision-making and action. In contemporary organizations, managers are expected not merely to access dashboards but to use evidence to allocate resources, manage performance, identify risks, improve customer value and guide digital transformation. The central managerial question is therefore not “How many reports does the organization produce?” but “Does information improve the quality, speed, fairness and consequences of decisions?”

Digital transformation is the coordinated redesign of strategy, processes, technology, structures and capabilities through digital technologies. It is broader than digitization, which converts analogue information into digital form, and digitalization, which uses digital technologies to improve an existing process. Transformation changes how an organization creates and captures value. BI supports this transformation by providing visibility into current performance, evidence for experimentation and feedback about whether change is producing value.

This chapter examines BI as a managerial and organizational capability. It addresses evidence-based management, performance management, BI maturity, operating models, adoption barriers, change management, data literacy, value realization and return on investment (ROI). It also considers governance, ethics, artificial intelligence (AI) and the limits of data-driven decision-making.

Learning Objectives

By the end of this chapter, students should be able to:

1. Explain how BI supports evidence-based managerial decision-making.
2. Distinguish reporting, diagnostic analysis, predictive analysis and prescriptive decision support.
3. Evaluate BI maturity across technology, governance, people and organizational practice.
4. Compare centralized, decentralized and federated BI operating models.
5. Diagnose adoption barriers and design suitable change-management responses.
6. Develop data-literacy and performance-management practices.
7. Construct a defensible BI business case and evaluate value realization and ROI.
8. Assess ethical, regulatory and organizational risks associated with data-driven management.

1. BI as a Management Capability

A BI system combines data sources, data integration, storage, analytical methods, visualization, governance and user practices. A dashboard is only the visible interface. The organizational capability includes the definitions, processes, skills and accountability that determine whether the displayed information is trusted and acted upon.

BI creates value through a chain:

Data → Information → Insight → Decision → Action → Outcome → Learning

Data consists of recorded observations, such as transactions or service times. Information is data interpreted in context, for example monthly sales by region compared with target. Insight is a meaningful explanation or implication, such as a decline caused by stock-outs in a particular channel. A decision selects a course of action. Action changes resource allocation or behaviour. Outcomes are the resulting effects, and learning updates future decisions, definitions and models.

The chain can fail at any point. Poorly governed data produces misleading information; attractive visualizations can obscure uncertainty; managers may ignore an accurate insight because incentives conflict with it; and an action may not be monitored, preventing learning. Consequently, BI should be managed as a sociotechnical system involving people, processes and technology.

Evidence-based management

Evidence-based management does not mean that managers follow a dashboard mechanically. It means that decisions are informed by the best available evidence, professional expertise, organizational values and the interests of affected stakeholders. Evidence may include internal operational data, external benchmarks, research findings, experiments, customer feedback and expert judgment.

A sound decision process normally includes:

1. Define the decision and its success criteria.
2. Identify relevant evidence and its quality.
3. Establish the baseline and comparison group where possible.
4. Examine alternative explanations and uncertainties.
5. Select and implement an action.
6. Monitor outcomes and revise the decision when evidence changes.

This approach guards against common errors such as confusing correlation with causation, relying on a single performance indicator, using averages that conceal subgroup effects and treating historical patterns as permanent rules.

2. From Descriptive Reporting to Decision Intelligence

BI applications can be classified by the question they answer.

Descriptive reporting is essential for visibility but does not by itself explain performance. Diagnostic analysis may compare actual results with budget, prior period, peer group or target. Predictive models estimate future outcomes under stated assumptions. Prescriptive systems evaluate possible actions, often subject to constraints such as budget, capacity or regulatory requirements.

The analytical level must match the decision. A stable operational process may need real-time alerts, whereas a strategic investment requires scenario analysis and sensitivity testing. More sophisticated analytics are not automatically better: a complex model with unreliable data may be less useful than a transparent descriptive measure.

Performance measurement and KPIs

A key performance indicator (KPI) is a measure linked to an important objective and accompanied by a defined interpretation and action. A metric becomes a KPI only when its strategic relevance, owner, target, time period and decision consequence are clear.

A useful KPI specification includes:

• name and business purpose;
• precise numerator and denominator, if applicable;
• population, inclusion and exclusion rules;
• data source and refresh frequency;
• target, threshold and tolerance;
• responsible owner;
• required action when performance deviates;
• known limitations and possible unintended effects.

For example, “customer satisfaction” is incomplete unless the organization specifies the survey population, response scale, sampling method and reporting period. “On-time delivery rate” requires a definition of on-time, the order population and treatment of cancelled or partially fulfilled orders.

Managers should balance outcome measures and driver measures. Revenue is an outcome; conversion rate, availability and sales response time may be drivers. Financial measures should be complemented by customer, process, people, risk and sustainability measures. A narrow target can create gaming or dysfunctional behaviour. If a call centre rewards short handling time alone, employees may terminate calls prematurely or transfer difficult cases.

Useful performance calculations

A variance compares an observed value with a reference value. The absolute variance is:

The percentage variance is:

Here, Actual and Reference must use the same unit and period, and Reference must not be zero. The sign must be interpreted in context: a positive cost variance may be unfavourable, while a positive revenue variance may be favourable.

For a rate such as conversion rate:

The numerator must be a subset of the denominator, and the observation window and eligibility rules must be consistent.

For an illustrative example, suppose a service target is 92% and actual performance is 88%. The percentage variance relative to target is:

The result indicates performance is approximately 4.35% below target, not that performance fell by 4 percentage points. The absolute difference is four percentage points; the relative variance is approximately −4.35%.

3. BI Maturity Assessment

BI maturity describes how consistently an organization converts data into trusted decisions and measurable outcomes. Maturity is multidimensional, not a simple technology ladder. An organization may have advanced cloud infrastructure but weak governance and low adoption.

A practical maturity model contains five stages:

1. Fragmented reporting: spreadsheets and isolated departmental reports dominate; definitions conflict and reporting is reactive.
2. Standardized reporting: common data sources, core KPIs and scheduled dashboards emerge, although analysis remains mainly descriptive.
3. Managed BI: governance, data ownership, service levels, self-service controls and data-quality monitoring are established.
4. Analytical organization: predictive and scenario analysis influence planning; experimentation and data literacy become routine.
5. Data-driven transformation: data products, embedded analytics, automation and continuous learning are integrated into operating and strategic models.

The stages are indicative rather than universal. A regulated organization may deliberately restrict self-service access while achieving high maturity through strong controls. Maturity should be assessed against strategic needs, risk tolerance and decision complexity.

Assessment dimensions

An assessment should combine interviews, artifact review, usage data, data-quality results, survey evidence and performance outcomes. A maturity score alone is insufficient. The assessment must identify the next capability that will improve important decisions.

4. BI Operating Models

The operating model defines who owns data, platforms, analytical standards, products and decisions. Three broad models are common.

Centralized model

A central BI or analytics function owns the platform, data models, standards and most delivery. Advantages include consistency, economies of scale, stronger governance and reduced duplication. Risks include slow response to business needs, a delivery queue and limited domain ownership.

Decentralized model

Business units own their data products and analytical teams. This can provide speed, domain expertise and strong local adoption. However, it often produces inconsistent definitions, duplicated technology, fragmented security and difficulty integrating enterprise-wide views.

Federated or hub-and-spoke model

A central team provides architecture, governance, shared platforms and enablement, while domain teams develop governed products with business ownership. This model seeks to combine enterprise consistency with local responsiveness. It requires explicit decision rights, common standards and collaboration mechanisms; otherwise it can become decentralization without coordination.

Operating-model design should specify:

• who defines enterprise metrics;
• who owns source-data quality;
• who approves access and classification;
• who funds shared platforms and analytical products;
• who supports users;
• who accepts model and compliance risk;
• how priorities are selected and conflicts resolved.

A responsibility-assignment matrix, such as RACI, can clarify whether a party is Responsible, Accountable, Consulted or Informed. Accountability should not be assigned only to IT: business owners understand meaning and consequences, while technical teams manage implementation and controls.

5. Adoption, Change and Data Literacy

Many BI programmes fail despite technically sound products. The causes often include low trust, inadequate training, poor workflow integration, competing incentives, executive inconsistency and the perception that analytics increases surveillance or threatens professional autonomy.

Adoption means more than logging into a dashboard. Meaningful adoption occurs when the right users consult relevant information, interpret it correctly, incorporate it into decisions and change behaviour where appropriate.

Diagnosing adoption barriers

Barriers may be grouped as follows:

• Capability barriers: users lack statistical, digital or domain skills.
• Access barriers: systems are slow, unavailable, poorly integrated or difficult to use.
• Trust barriers: data definitions, quality or model outputs are disputed.
• Meaning barriers: measures are disconnected from roles and decisions.
• Incentive barriers: targets reward local optimization or discourage transparency.
• Political barriers: information changes power relationships or exposes weak performance.
• Ethical barriers: employees or customers fear inappropriate monitoring or discrimination.

Change-management approach

A practical change programme should begin with a compelling decision problem, not a technology launch. Leaders should communicate why the change matters, involve affected users in design, provide role-specific training and create mechanisms for feedback. Early pilots should address high-value, manageable use cases and demonstrate both benefits and limitations.

A dashboard should be embedded in a management ritual: a weekly service review, monthly performance meeting, sales pipeline discussion or quarterly planning cycle. The meeting should specify who examines which indicators, what questions are asked, what actions are recorded and when results are reviewed. Without this operating rhythm, dashboards become passive information displays.

Data literacy is the ability to read, understand, question, communicate and use data responsibly. It includes vocabulary, measurement concepts, interpretation of charts, basic statistical reasoning, awareness of bias, privacy principles and the ability to connect evidence with decisions. Data literacy should be role-based. A senior executive needs to interpret uncertainty and trade-offs; a frontline employee may need to understand alerts and data-capture rules; an analyst needs stronger methodological and governance skills.

Effective development includes short practical learning sessions, annotated examples, office hours, communities of practice, data dictionaries and coaching during real decisions. Training should be evaluated through task performance and decision quality, not attendance alone.

6. Value Realization and ROI

A BI initiative should be justified by the decisions and outcomes it enables. Potential benefits include increased revenue, reduced cost, lower risk, improved service, faster cycle time, regulatory compliance and better employee or customer experience. Some benefits are directly measurable; others require proxies or qualitative assessment.

A value-realization plan should define:

1. the baseline before intervention;
2. the target outcome and measurement period;
3. the mechanism by which BI is expected to create value;
4. the owner of each benefit;
5. implementation and recurring costs;
6. assumptions and external influences;
7. the review schedule and corrective actions.

Common costs include software, infrastructure, implementation, data engineering, migration, training, support, governance and the time users spend changing processes.

A simple ROI measure is:

Benefits and costs must be measured over the same time horizon and expressed in comparable monetary units. If benefits occur over several years, a discounted cash-flow approach is more appropriate. Net present value (NPV) is:

Here, C₀ is the initial investment at time zero, CFₜ is the net cash flow in period t, r is the discount rate per period, and n is the number of periods. The formula assumes cash flows are assigned to defined periods and that the discount rate reflects the relevant opportunity cost and risk. A positive NPV indicates value relative to the chosen discount rate, but it does not prove that the project is strategically or ethically desirable.

Worked ROI example

An organization invests £120,000 in a governed performance-management solution. During the first year it records £165,000 in verified benefits and £30,000 in operating costs. Total costs are £150,000, and net benefits are £15,000.

The result is a simple first-year ROI of 10%. The calculation does not establish causation automatically. The organization should verify whether savings were actually realized, whether revenue effects were incremental, whether benefits were double-counted and whether the baseline was credible.

Value realization should also track leading indicators, such as active use by decision-makers, data-quality improvement, time saved in reporting and the percentage of management actions linked to evidence. These indicators do not replace outcome measures, but they help explain why benefits have or have not appeared.

7. Governance, Ethics and Responsible Use

BI governance establishes decision rights, standards and controls across the data life cycle. Important components include data ownership, metadata, quality management, access control, retention, privacy, security, model governance and issue escalation.

Data quality is multidimensional. Accuracy concerns whether data reflects reality; completeness concerns missing values or records; consistency concerns agreement across systems; timeliness concerns availability when needed; validity concerns conformity with permitted formats or rules; and uniqueness concerns duplicate records. Quality is fitness for a specific use, not an abstract perfection score.

Ethical management requires attention to:

• lawful and fair collection and use of personal data;
• purpose limitation and data minimization;
• transparency about automated decisions;
• access, correction and deletion rights where applicable;
• protection against unauthorized disclosure;
• testing for discriminatory outcomes;
• human review of high-impact decisions;
• explainability appropriate to the affected audience;
• retention and secure disposal;
• accountability for vendor and AI-system behaviour.

A model can be statistically accurate overall yet unfair to a subgroup. Managers should examine performance across relevant groups, document trade-offs and avoid using sensitive or proxy variables without a legitimate, carefully assessed purpose. Explainability is not merely a technical feature; it is also the ability to communicate what data was used, what the output means, how uncertain it is and how a person can challenge it.

AI and generative AI can accelerate natural-language querying, forecasting, anomaly detection and narrative reporting. They also introduce risks such as hallucinated explanations, leakage of confidential information, automation bias, unstable outputs and hidden changes in model behaviour. AI-generated analysis should be treated as an input requiring verification, not as an authoritative source. Production use requires approved data, access controls, logging, testing, human accountability and monitoring for drift.

8. Digital Transformation Through BI

BI contributes to transformation in four connected ways. First, it provides transparency about the current operating model. Second, it identifies bottlenecks and customer needs. Third, it supports experimentation and rapid feedback. Fourth, it embeds intelligence into processes, products and services.

Transformation should not begin with a catalogue of technologies. It should begin with strategic questions such as:

• Which customer or operational problem requires redesign?
• What decision is currently slow, inconsistent or poorly informed?
• What capabilities must change, and who owns them?
• What data is required, and is its use legitimate?
• How will the organization know that transformation created value?

Examples include predictive maintenance linked to work-order scheduling, personalized recommendations connected to customer-service capacity, or supply-chain visibility linked to procurement decisions. In each case, insight must be connected to action. A forecast that is not integrated into planning may have little practical value.

Transformation also changes organizational power. Greater transparency may challenge established assumptions, expose process variation or shift authority from hierarchy toward expertise and evidence. Senior leaders must therefore model responsible challenge, tolerate learning from failed experiments and align incentives with enterprise rather than narrow local outcomes.

9. Illustrative Case: Northstar Retail Group

Background

Northstar Retail Group operates 80 stores and an online channel. It has invested in a cloud data platform and executive dashboards, but regional managers continue to maintain spreadsheets. Sales, inventory and promotion definitions differ between departments. Stock availability is declining, while inventory investment is increasing.

Problem

The executive team initially assumes that a more advanced forecasting algorithm will solve the problem. A diagnostic review shows that the main barriers are inconsistent product hierarchies, delayed stock updates, weak ownership of inventory data and meetings that focus on explaining last month rather than deciding future replenishment.

Analysis

The organization maps the decision chain from demand signal to purchase order. It identifies three critical decisions: replenishment quantity, promotion approval and allocation of scarce stock. Baseline measures include stock-out rate, inventory value, forecast error, gross margin and the time required to prepare weekly reports. The team also identifies the risk that managers may over-order to protect local availability, shifting costs to the enterprise.

Approach and solution

Northstar adopts a federated model. A central team owns common definitions, security, platform architecture and data-quality rules. Merchandising and supply-chain teams own governed analytical products for their domains. A data council approves enterprise definitions and resolves conflicts.

The organization introduces a balanced performance view:

• stock-out rate and availability;
• inventory days and working capital;
• gross margin and markdown rate;
• forecast error by product category;
• supplier lead-time reliability;
• customer complaints and lost-sales estimates.

Managers receive role-specific dashboards. The weekly meeting begins with exceptions and decisions rather than presentation of every metric. Each action has an owner, expected effect and review date. A pilot is conducted in 12 stores, with comparison against similar stores not yet using the redesigned process.

The organization also introduces data-literacy workshops on definitions, uncertainty and causal interpretation. The forecast is presented with confidence intervals and scenario assumptions rather than as a single unquestionable number. No automated replenishment recommendation is implemented without a human review threshold.

Outcome

After the pilot, Northstar reports improved availability and lower emergency transfers. It does not attribute all improvement to BI: seasonal demand, supplier changes and management attention may also contribute. The pilot therefore compares results with the baseline and comparison stores and documents the limitations. The organization scales the approach only after confirming that benefits persist and that no region is systematically disadvantaged by the new allocation rules.

Learning points

The case illustrates that transformation depends on definitions, decision rights, routines and adoption as much as on algorithms. A technically sophisticated forecast cannot compensate for unreliable data or an unchanged operating process. Value is realized when insight is connected to accountable action and evaluated against a credible baseline.

10. Challenges and Limitations

BI can create false confidence. Historical data may reflect past discrimination, exceptional events or obsolete processes. Aggregated indicators can conceal variation between customers, locations or employee groups. Real-time data can accelerate poor decisions if speed is valued over reflection. Too many KPIs create attention fragmentation, while too few encourage oversimplification.

Causal attribution is another difficulty. If performance improves after a dashboard launch, the improvement may result from market conditions, leadership attention, staffing or unrelated initiatives. Where feasible, organizations should use controlled pilots, comparison groups, interrupted time-series analysis or triangulation with qualitative evidence. No evaluation design eliminates all uncertainty, but explicit assumptions make conclusions more credible.

There are also trade-offs between standardization and flexibility, privacy and personalization, transparency and security, automation and human judgment, and short-term measurable benefits and long-term capability building. Senior managers must make these trade-offs explicit rather than treating technology as neutral.

Best-Practice Checklist

A responsible BI and transformation programme should:

1. Start with important decisions and strategic outcomes.
2. Establish common definitions and accountable data owners.
3. Assess maturity across technology, people, governance and culture.
4. Select an operating model with clear decision rights.
5. Design dashboards around actions, not visual variety.
6. Embed analytics in management routines and workflows.
7. Build role-specific data literacy.
8. Measure adoption, outcomes and unintended consequences.
9. Use credible baselines and avoid overstating causation.
10. Protect privacy, security, fairness and the right to challenge decisions.
11. Govern AI outputs through testing, monitoring and human accountability.
12. Reinvest in the capabilities that produce sustained value.

Chapter Summary

BI is an organizational capability that links data to insight, decisions, actions and learning. Its managerial value depends on trustworthy definitions, suitable analytical methods, clear accountability and integration into work practices. Descriptive, diagnostic, predictive and prescriptive analytics answer different questions and should be selected according to decision needs.

Performance management requires balanced KPIs with explicit definitions, targets, owners and consequences. BI maturity includes strategy, data, technology, governance, people, process, culture and value; advanced technology alone does not indicate maturity. Centralized, decentralized and federated operating models each provide different trade-offs between consistency, responsiveness and domain ownership.

Adoption depends on trust, capability, incentives, workflow integration and leadership. Data literacy enables responsible interpretation, while change management turns dashboards into decision routines. Value realization requires a baseline, benefit owners, verified outcomes and transparent assumptions. ROI is useful but should be complemented by NPV, adoption measures, risk assessment and strategic evaluation.

Digital transformation is achieved when data and analytics redesign how value is created, not merely when new tools are installed. Governance, privacy, fairness, security and human accountability are essential, especially when AI is used. The most effective organizations treat BI as a continuous learning system in which evidence informs action and results improve future decisions.

Key Terms

Business intelligence: Organizational capability for collecting, integrating, analyzing and using data to support decisions.

Evidence-based management: Decision-making informed by the best available evidence, expertise, values and stakeholder interests.

Digital transformation: Redesign of strategy, processes, capabilities and value creation through digital technologies.

Digitization: Conversion of analogue information into digital form.

Digitalization: Use of digital technology to improve an existing process.

KPI: A strategically important measure linked to a target, owner and management action.

Descriptive analytics: Analysis concerned with what has happened.

Diagnostic analytics: Analysis concerned with why an outcome occurred.

Predictive analytics: Analysis that estimates likely future outcomes.

Prescriptive analytics: Analysis that evaluates or recommends possible actions.

Data literacy: Ability to understand, question, communicate and responsibly use data.

BI maturity: The level of capability with which an organization turns data into trusted decisions and outcomes.

Operating model: Arrangement of roles, responsibilities, governance and resources used to deliver BI.

Federated BI: Model combining central standards and platforms with domain-level analytical ownership.

Data governance: Decision rights, policies, standards and controls for managing data.

Data quality: Fitness of data for a specified purpose, including accuracy, completeness and timeliness.

Value realization: Process of tracking whether expected benefits from an initiative are achieved.

Return on investment: Net benefits divided by total costs, expressed as a percentage.

NPV: Present value of future net cash flows minus the initial investment.

Automation bias: Tendency to place excessive trust in an automated output.

Review Questions

Short Answer Questions

1. What distinguishes digital transformation from digitization?
2. Why is a dashboard not, by itself, a BI capability?
3. Define a KPI and identify three elements that should accompany it.
4. What is data literacy, and why is it important for managers?
5. State two ethical risks associated with AI-supported BI.

Long Answer Questions

1. Explain how descriptive, diagnostic, predictive and prescriptive analytics support different managerial decisions.
2. Compare centralized, decentralized and federated BI operating models, including their principal advantages and risks.
3. Discuss the dimensions of BI maturity and explain why technology maturity alone is insufficient.
4. Explain how change management, management routines and incentives influence BI adoption.
5. Evaluate the use of ROI and NPV for assessing a BI investment, including their assumptions and limitations.

Analytical / Application Questions

1. A company reports actual customer retention of 84% against a target of 90%. Calculate the percentage variance relative to target and explain why the result should not be described simply as a six-percent decline.
2. Design a BI maturity assessment for a hospital, including at least six dimensions, evidence sources and recommended next steps.
3. A retailer wants to automate credit decisions using AI. Propose a governance and value-realization plan that addresses data quality, fairness, human review, privacy, adoption and outcome measurement.

Multiple Choice Questions with Answers

Question 1

Which statement best describes evidence-based management?

A. Managers follow dashboards without using professional judgment.

B. Managers use the best available evidence alongside expertise, values and stakeholder interests.

C. Managers rely only on historical financial data.

D. Managers replace all qualitative information with algorithms.

Correct Answer: B. Managers use the best available evidence alongside expertise, values and stakeholder interests.

Question 2

Which analytical level primarily addresses the question “Why did it happen?”

A. Descriptive analytics

B. Diagnostic analytics

C. Predictive analytics

D. Prescriptive analytics

Correct Answer: B. Diagnostic analytics

Question 3

Which feature is most characteristic of a federated BI operating model?

A. All analysis is performed by one central department.

B. Every department uses independent definitions and platforms.

C. Central standards coexist with analytical ownership in business domains.

D. No formal governance is required.

Correct Answer: C. Central standards coexist with analytical ownership in business domains.

Question 4

Actual performance is 88% and the target is 92%. What is the approximate percentage variance relative to target?

A. −4.00%

B. −4.35%

C. 4.35%

D. −4 percentage points only

Correct Answer: B. −4.35%

Question 5

Which practice most directly supports meaningful dashboard adoption?

A. Increasing the number of charts on every page

B. Embedding dashboard review in a recurring management decision routine

C. Restricting all dashboards to the IT department

D. Removing targets so users cannot be challenged

Correct Answer: B. Embedding dashboard review in a recurring management decision routine

Question 6

In a simple ROI calculation, benefits are £165,000 and total costs are £150,000. What is the ROI?

A. 10%

B. 15%

C. 90.9%

D. 110%

Correct Answer: A. 10%

Question 7

Which statement about data quality is most accurate?

A. Data quality means data is perfect for every possible use.

B. Data quality is fitness for a specified purpose and includes dimensions such as accuracy and timeliness.

C. Data quality is determined only by the size of a database.

D. Data quality is unrelated to business definitions.

Correct Answer: B. Data quality is fitness for a specified purpose and includes dimensions such as accuracy and timeliness.

Question 8

Why should a BI business case include a credible baseline?

A. To guarantee that the initiative will be profitable

B. To help assess whether observed changes are plausibly attributable to the initiative

C. To eliminate the need for governance

D. To ensure that all benefits are qualitative

Correct Answer: B. To help assess whether observed changes are plausibly attributable to the initiative

Question 9

Which is an example of automation bias?

A. A manager challenges an anomalous model output.

B. A user refuses to use any digital system.

C. A decision-maker accepts an automated recommendation without appropriate review.

D. An analyst documents model assumptions.

Correct Answer: C. A decision-maker accepts an automated recommendation without appropriate review.

Question 10

What is the primary purpose of a data council in a federated BI model?

A. To replace every business analyst

B. To coordinate definitions, standards and cross-domain governance

C. To prevent business users from accessing data

D. To approve marketing campaigns only

Correct Answer: B. To coordinate definitions, standards and cross-domain governance