AnalyticsArticleNovember 18, 2026
Data Governance for Analytics: Who Owns the Number Data governance for analytics is the set of working rules that decide who owns each metric, how its definition is agreed, who can see it, and how it changes.

Data Governance for Analytics: Who Owns the Number

Data Governance for Analytics: Who Owns the Number Data governance for analytics is the set of working rules that decide who owns each metric, how its definition is agreed, who can see it, and how it changes.

Sophia Moreau
Sophia Moreau
22 min read

Data Governance for Analytics: Who Owns the Number

Data governance for analytics is the set of working rules that decide who owns each metric, how its definition is agreed, who can see it, and how it changes. When those rules exist and people follow them, two dashboards show the same revenue number. When they don’t, every team builds its own version of the truth, and leadership stops believing any of them. At Ridiculous Engineering, we treat this as an operating model problem first and a tooling problem second.

Ridiculous Engineering
ridiculousengineering.com
Build Trust in Your Analytics
We help organizations connect data systems, develop analytics, and establish practical technology solutions around their business needs.
Explore our services

Key Takeaways

  • Distrust in dashboards usually comes from unowned definitions, not bad data. Name one accountable owner for every metric that reaches an executive.
  • Governance for analytics comes down to four rules: ownership, agreed definitions, access, and change control. Each can be run with a small group and a shared document before any platform purchase.
  • Stewardship works when business stewards own meaning and technical stewards own pipelines. Splitting the two keeps definitions from being decided by whoever writes the SQL.
  • Measure governance like any operating process: share of metrics with an owner and a definition, share of data meeting quality standards, and how much notice users get before a metric changes.
  • Start with the five to ten metrics that appear in board and leadership reviews, then widen the scope.

Table of Contents

The symptom: two dashboards, two different revenue numbers

Picture a Monday leadership meeting. Finance shows quarterly revenue from the billing system. Sales shows a figure from the CRM that is several percent higher. Operations shows a third number from the order system. Nobody is lying, and nobody’s query is broken.

Finance counts revenue when an invoice is issued. Sales counts it when a contract is signed. Operations counts it when an order ships. Refunds, multi-year deals, currency conversion and credit notes are each handled differently, and each team made a reasonable choice at the time. The problem is that nobody decided which choice is the company’s choice.

The cost is not just the argument itself. Meetings turn into reconciliation sessions. Analysts spend their week explaining differences instead of finding insight. Leaders fall back on spreadsheets and gut feel, and the dashboard investment sits unused. Forrester’s analysis, published in Forbes, found that fewer than 10% of enterprises are advanced in insights-driven capabilities, and it points to organizational habits and practices, not technology, as what separates beginners from advanced organizations. Those habits and practices are what data governance is.

This is the pattern we see described again and again in analytics programs: the tools work, the numbers disagree, and the missing piece is a decision nobody was assigned to make.

The symptom: two dashboards, two different revenue numbers

What data governance for analytics actually is

Strip away the vocabulary and governance answers four plain questions about every number people use to make decisions:

  1. Who is accountable for it?
  2. What exactly does it mean?
  3. Who is allowed to see and use it?
  4. What happens when it needs to change?

Gartner frames data governance as the specification of decision rights and an accountability framework for how data and analytics are valued, created, consumed and controlled. The phrase “decision rights” is the useful part. Governance is less about documents and more about who gets to decide, and what they must do before they decide it.

Gartner’s guidance for finance leaders puts the purpose in a similar way: governance should support decision makers and drive value, not exist as a standalone control function. If your governance effort produces a policy binder that analysts never open, it is not doing that job.

Governance for analytics versus governance for compliance

Privacy and regulatory obligations matter, and your legal team has its own program for them. This article is about something narrower and more daily: the rules that make a metric mean the same thing in every report. A company can be fully compliant and still have three revenue numbers. The reverse is also true. Treating the two as one project is how analytics governance ends up stalled in a legal review queue.

The four working rules

The federal Data Governance Playbook lists six core activities: identifying data through inventory and metadata, setting data management policy, managing data issues, assessing data quality and utility, providing oversight, and communicating. That is a sound outline for a government agency. For a mid-size company that wants consistent analytics, those activities collapse into four rules you can start running within weeks.

Rule 1: Every metric has one named owner

An owner is a person, not a team or a committee. They are accountable for the metric’s definition, for approving changes, and for answering when someone questions the number. For revenue, that is usually a finance leader. For customer churn, it may be a head of customer success. For on-time delivery, an operations leader.

A good owner is someone in the business who cares about the number being right, not whoever built the first dashboard. If you can’t name the owner of a metric that appears in executive reviews, you have found your first governance gap.

Rule 2: Definitions live in one place, in plain language

A metric definition needs more than a name. At minimum, write down:

  • The business meaning in one or two sentences.
  • The calculation, including what is included and excluded (refunds, internal accounts, test orders).
  • The source system the number comes from.
  • The time basis: when an event counts, and in which time zone and fiscal calendar.
  • The owner and the date of the last approved change.

This is often called a metric dictionary or business glossary. The format matters less than two properties: there is one copy, and the dashboards actually draw from it. A glossary that lives in a wiki while dashboards use hand-written SQL will drift within a quarter.

When two teams define a metric differently for good reasons, don’t force one definition onto everyone. Give each a distinct name. “Booked revenue” and “recognized revenue” can both exist. “Revenue” on its own should mean exactly one of them.

Rule 3: Access follows the data’s sensitivity and the job

Analytics access is where governance meets daily frustration. Lock everything down and analysts build shadow copies in spreadsheets. Open everything up and sensitive compensation, customer or contract data ends up in a widely shared workbook.

A workable approach is to classify data into a few tiers, such as general, internal, and restricted, and grant access by role to each tier. Keep the number of tiers small. People follow three levels; they ignore nine. Record who approved each exception, and review the list on a schedule so access doesn’t accumulate as people change jobs.

The same discipline carries over as teams add AI-driven search and question-answering on top of internal data: the permissions model decides what an assistant can retrieve, so it has to be sound before the assistant is connected. Platforms for AI search over internal data show how permissions and audit trails become part of the product itself once answers are generated from many sources.

Rule 4: Changes are controlled and announced

Most trust is lost not when a number is wrong, but when it quietly changes. A revenue figure shifts by 2% after someone fixes a join, and nobody told the people who quote it to the board.

Treat metric definitions and the pipelines behind them the way engineering teams treat code:

  • Proposed changes are written down with the reason and the expected effect on historical numbers.
  • The metric owner approves them.
  • Affected report users are told in advance. The FHWA’s data governance primer, for example, uses a sample measure of affected users knowing about data changes at least 30 days before implementation. Thirty days is one reasonable standard; pick the window that fits your reporting cycle.
  • The change is versioned, so anyone can see what the definition was last quarter.

When a restated number does reach a leadership review, the owner explains it before anyone has to ask.

Stewardship roles: who does the work

Owners decide. Stewards keep the day-to-day work moving. The research on stewardship consistently separates two kinds of steward, and the split is worth copying.

In her book on designing data governance, Lauren Maffeo describes business stewards as people who review taxonomies, write documentation, lead governance training and run quality assessments against predefined business metrics. Technical stewards, by contrast, manage the systems and pipelines, including warehouses and lakes, and define metadata parameters.

In practice for a mid-size company:

Role Owns Typical person Example question they answer
Metric owner The decision on what a metric means and when it changes Functional leader (finance, sales, operations) “Should we count paused subscriptions as active customers?”
Business steward Definitions, documentation, quality checks against business rules Senior analyst or operations manager “Is the glossary entry current, and do the dashboards match it?”
Technical steward Pipelines, source systems, metadata, access configuration Data or analytics engineer “Why did this table load late, and who is affected?”
Governance lead The cadence, the backlog, escalation Head of BI or analytics “Which disputes are unresolved, and which metrics lack owners?”

The FHWA primer describes data stewards as subject matter experts ultimately responsible for defining, managing and preserving the integrity of data. Note the word “subject matter experts”. Stewardship is a role added to an existing job, not a new department. Most mid-size companies assign it to people who already know the data, give them a few hours a week, and make the responsibility explicit in their goals.

The split between business and technical stewards also prevents a common failure: definitions being decided, by default, by whoever writes the query. Engineers are good at making a calculation run. They shouldn’t be the ones deciding whether a cancelled order counts as revenue.

A small governance council

You also need a place where disputes get settled. A monthly or biweekly council of metric owners, the head of BI and a technical steward is usually enough. The federal playbook argues that a governance body must be integrated into decision-making and operations to be effective, and that executives should receive regular updates on its impact. Translated to a company: put governance on the leadership agenda with a short, recurring report, or it will be treated as optional.

Data quality: define it, then measure it

Quality is where governance becomes measurable. The data quality framework most teams search for boils down to three steps: define what “good” means for each critical dataset, measure it on a schedule, and assign someone to fix what fails.

GS1 US guidance says organizations can’t manage what they can’t measure, and recommends measuring completeness, timeliness and accuracy with clearly defined metrics, valid ranges for first-time and ongoing monitoring, and baselines to track progress. A baseline matters because it lets you show improvement instead of asserting it.

A practical starting set of checks for any dataset behind a leadership metric:

  • Completeness: Are required fields populated? What percent of orders lack a customer ID?
  • Timeliness: Did the nightly load finish before the morning report refreshed?
  • Accuracy: Do totals reconcile with the source system within an agreed tolerance?
  • Consistency: Does the same customer count match between the CRM and the warehouse?

Each check needs a threshold, an owner, and an action when it fails. Gartner’s analytics strategy guidance adds a point that helps with budget conversations: create metrics for both the value of high-quality data and the cost of low-quality data, and place people with data quality skills in the parts of the business where poor data costs money. For a deeper walk-through of the technical side, see our practical guide to data quality management.

Starting small: a 90-day plan

Gartner lists seven pillars for governance that supports business outcomes, and the first is starting with business outcomes rather than with data inventory. That is the right order for analytics leaders. Here is a plan that follows it.

  1. Weeks 1 to 2: Pick the metrics. List the five to ten numbers that appear in board, executive and weekly operating reviews. Ignore everything else for now.
  2. Weeks 2 to 4: Name owners and write definitions. For each metric, record the owner, definition, source, and time basis. Where two definitions exist, rename them and decide which one gets the plain name.
  3. Weeks 4 to 6: Find the divergence. Compare each metric across the dashboards that show it. Document every gap and trace it to a definition difference, a data problem, or a timing difference.
  4. Weeks 6 to 8: Point dashboards at the shared definition. Move calculations out of individual reports and into one governed layer, so the logic is written once. This is the step that usually needs engineering effort, and it is where governance stops being a document.
  5. Weeks 8 to 10: Set access tiers and change rules. Classify the datasets behind those metrics, set role-based access, and start announcing changes in advance.
  6. Weeks 10 to 12: Start the cadence. Hold the first council meeting, publish the first short scorecard, and choose the next batch of metrics.

The constraint is deliberate. A governance program that tries to catalog every table before delivering anything tends to lose its sponsor. One that fixes the revenue number first earns credibility for the second phase.

Centralized, federated, or somewhere between

Companies often ask whether one central team should own everything or whether each department should govern its own data. For mid-size organizations, a hybrid is usually the practical answer: a small central group sets the standards, the glossary and the access tiers, while functional leaders own their metrics and stewards sit in the business. Central control alone becomes a bottleneck, and departmental control alone brings back the two-revenue-numbers problem.

Hybrid model linking central standards with local ownership

How to tell whether governance is working

The federal playbook advises developing metrics to assess whether the data vision is being realized, collecting the data to inform them, acting on results, and periodically assessing data maturity to prioritize next steps. For an analytics team, a small scorecard is enough:

Measure What it tells you
Share of executive metrics with a named owner and written definition Whether accountability exists
Number of metrics that show different values in different reports Whether definitions are truly shared
Share of critical datasets passing quality checks Whether the data behind the metrics is dependable
Average days of notice before a metric changes Whether change control is real
Open definition disputes and their age Whether the council is resolving things
Hours analysts spend reconciling numbers The cost you started with, going down

The FHWA primer offers similar sample measures, such as the percent of data elements with clear definitions and the percent of data meeting accuracy standards. Record the baseline in week one so the trend means something. The last measure on the list is the one executives understand fastest: time spent arguing about numbers instead of acting on them.

Where engineering meets governance

Governance rules fail when they depend on people remembering them. The durable version puts the rules into the system: a single governed layer for metric logic, access rules enforced by the platform, automated quality checks that run with each load, and version history for every definition change. That is software engineering work, and it is where policy turns into behavior.

Our Data Analytics and Business Intelligence service covers data analytics systems and business intelligence. Clients work directly with senior engineers who build, modernize and integrate practical systems around how the organization actually operates, with support from discovery through long-term improvement. If your dashboards disagree and you want the definitions, access rules and change controls built into the systems that produce them, talk to us about a scoped engagement. Pricing is based on scope, so we quote it on request.

FAQ

What is data governance for analytics in plain terms?

It is the set of rules that decides who owns each metric, how its definition is agreed, who can see the underlying data, and how changes are made. The goal is that the same question gets the same answer in every report. It is less about policy documents and more about assigning decisions to named people.

What is the difference between a data owner and a data steward?

An owner is accountable for a metric or dataset and makes the decisions about its meaning and changes. A steward does the ongoing work: maintaining definitions, running quality checks, and managing the pipelines and metadata. Business stewards focus on meaning and quality against business rules, while technical stewards focus on systems and pipelines.

Which data governance best practices should we start with?

Start with the handful of metrics leadership uses most. Name an owner for each, write one definition in plain language, and put the calculation in one shared place. Then add role-based access in a few tiers, announce changes before they happen, and track a small scorecard of ownership, definition coverage and quality.

How is a data quality framework different from data governance?

A data quality framework is one part of governance. It defines what good data means (for example completeness, timeliness and accuracy), measures it against thresholds, and assigns fixes. Governance is wider: it also covers ownership, definitions, access and change control, which determine whether quality checks are applied to the right things.

How long does it take to see results?

A focused effort on five to ten executive metrics can resolve the most visible disagreements within a quarter, since the work is mostly decisions and a shared calculation layer. Wider coverage follows in phases. The speed depends mostly on how quickly metric owners are named and definition disputes are settled.

Ridiculous Engineering
Discuss Your Data Governance Needs
Share the analytics challenges your organization is facing, and our team can discuss relevant engineering and technology approaches.

Sources

Embrace Technology with Confidence

Your Guide to Successful Technology Adoption

If you are looking for a guide in adopting technology, a technology switch, or how to best apply new technology in your business, we at Ridiculous Engineering are here for you. Reach out today to learn how we can help.