Key Performance Indicator (KPI)

A KPI is a quantifiable metric tied to a specific business objective - not just any number you can measure, but one that signals whether the organization is moving toward or away from a strategic goal. Revenue growth, customer churn rate, deployment frequency, net promoter score: these qualify as KPIs because someone changes behavior based on them.

The challenge is not defining KPIs but governing them. When finance calculates "monthly revenue" one way, marketing calculates it another, and the CEO's dashboard shows a third number, the metric stops being useful. That is a metadata problem, and it is where KPIs intersect with data governance.

What Makes a Good KPI

Five criteria separate a genuine KPI from a number that sits in a dashboard and gathers dust.

Tied to a decision. If no one changes behavior based on this number, it is not a KPI. "Total website visitors" is a metric. "Visitor-to-trial conversion rate" is a KPI - when it drops, the marketing team investigates and acts.

Clearly defined. The calculation method, data sources, and exclusions are documented. Monthly recurring revenue (MRR) can be a good KPI when it is defined this way: the monthly value of all active subscriptions, excluding one-time fees, taken from a named source system as the record of truth.

Consistently measurable. Measuring the same period twice, or by two different people, gives the same result. If the number depends on which spreadsheet someone pulled, or which dashboard they opened, it fails this test.

Owned. Someone is accountable for the metric and its trajectory. MRR is owned by finance. Customer churn is owned by customer success. If nobody owns a KPI, nobody acts on it.

Time-bound. Measured against a target within a specific period. "Reduce churn" is a wish. "Reduce monthly churn from 3.2% to 2.5% by Q4" is a KPI with a target. These criteria overlap with the SMART mnemonic (specific, measurable, assignable, realistic, time-related) that George T. Doran proposed for writing objectives in Management Review in 1981.

Leading vs. Lagging Indicators

Leading indicators predict future results. Lagging indicators confirm past performance. The distinction matters because organizations that only track lagging indicators are driving by looking in the rearview mirror.

Leading indicators signal what is likely to happen. Sales pipeline value predicts next quarter's revenue. Website traffic predicts conversion volume. Falling employee engagement scores can signal attrition risk before people actually leave. These metrics give teams time to act before the outcome is locked in.

Lagging indicators report what already happened. Quarterly revenue, annual profit margin, customer churn rate - these confirm whether the strategy worked. They are essential for accountability and reporting, but they arrive too late to change course.

The practical insight: every KPI dashboard should include both types. A SaaS company tracking only MRR (lagging) misses the pipeline signals that predict next month's MRR. A company tracking only pipeline value (leading) cannot confirm whether pipeline quality actually converts to revenue. Pairing them - pipeline value now, revenue next quarter - creates a feedback loop that validates or corrects assumptions.

The KPI Definition Problem

The most common KPI failure is not picking the wrong metric. It is having multiple conflicting definitions of the same metric.

Consider "active users." Marketing defines it as anyone who logged in within 30 days. Product defines it as anyone who performed a core action within 7 days. Finance defines it as any paid account with at least one session. Marketing reports 500,000 active users. Product reports 200,000. Finance reports 80,000. The CEO asks "how many active users do we have?" and gets three different answers.

This is not a measurement error. Each definition is internally consistent and reasonable for its context. The problem is that the organization never agreed on a single canonical definition. There is no governed business glossary entry for "active users" that specifies the calculation method, the data source, the inclusion/exclusion criteria, and the owner.

The fix starts with metadata rather than more meetings. KPI definitions belong in a governed glossary where the calculation, source, filters, and owner are documented and changes are reviewed. When someone changes the definition (say, switching from 30-day to 7-day active users), the change goes through review, the owners of downstream dashboards can be identified, and stakeholders can be told before the numbers move. Without this process, definition drift happens silently.

KPI Frameworks

Three frameworks help organizations structure their KPIs beyond ad-hoc selection.

Balanced Scorecard. Introduced by Robert Kaplan and David Norton in Harvard Business Review in 1992, this framework forces measurement across four perspectives: financial (revenue, margins), customer (satisfaction, retention), internal business (efficiency, quality), and innovation and learning, later usually called internal business process and learning and growth (employee development, innovation). The value is balance - it prevents the trap of optimizing financial KPIs while ignoring the customer and operational metrics that drive them.

OKRs (Objectives and Key Results). A qualitative objective ("Become the preferred vendor for mid-market SaaS companies") paired with 3-5 measurable key results ("Increase mid-market pipeline by 40%," "Win 15 new mid-market logos," "Achieve NPS 60+ in mid-market segment"). The method was created by Andy Grove at Intel and brought to Google by John Doerr in 1999. OKRs are common in technology companies and work well for goal alignment across teams. The risk: teams pick easy key results that look good but do not stretch performance.

North Star Metric. A single metric that captures the core value a product delivers. For a lodging marketplace that might be nights booked; for a music streaming service, time spent listening. The North Star anchors all team-level KPIs - every team's metrics should ultimately contribute to moving it. The weakness: a single metric can oversimplify complex businesses with multiple revenue streams.

Governing KPIs

Ungoverned KPIs degrade over time. Here is how to prevent it.

Central glossary. Every KPI has one governed definition, one owner, one calculation method. The glossary is the authoritative source. When a BI tool displays "Monthly Revenue," the calculation behind it should match the glossary definition - not a local formula someone built in a spreadsheet.

Lineage. Trace from the dashboard number back through transformations to source data. When the revenue number looks wrong, lineage shows which tables and transformations produced it, which shortens the investigation considerably.

Change management. When a definition changes (say, switching from 30-day to 7-day active users), the change is reviewed, the dashboards that use the metric are updated together, and stakeholders are told. Without this, one team changes the definition, the dashboard updates silently, and a month later someone notices the trend line "broke."

Quality monitoring. Automated checks that KPI values are within expected ranges. If "monthly revenue" suddenly drops to zero or jumps 500%, a data observability alert fires before the number reaches a dashboard. This catches pipeline failures and source data issues before they corrupt decision-making.

Common KPI Failures

Vanity metrics. Measuring total registered users instead of active users because the number is bigger. Tracking total page views instead of conversion rate because it always goes up. Vanity metrics are designed to impress, not to inform. They feel good in a board deck but do not drive decisions.

Gaming. Call center agents measured on average call duration may start cutting complex calls short. Sales reps measured on deal count may start splitting deals into smaller contracts. The pattern is often summed up as Goodhart's law, in anthropologist Marilyn Strathern's 1997 phrasing: "When a measure becomes a target, it ceases to be a good measure." KPIs shape behavior - if the metric incentivizes the wrong behavior, the team will optimize for the metric and harm the outcome.

Metric proliferation. An organization tracking 200 KPIs is rarely tracking any of them effectively. When everything is a priority, nothing is. Keep executive dashboards to a handful of KPIs and department dashboards to a short list; beyond that, the term "key" loses meaning.

Dashboard rot. Dashboards built for a specific initiative that nobody updates when business context changes. A year later, the dashboard still shows metrics for a product line that was discontinued. Users learn to distrust the entire analytics platform.

Definition drift. In a typical case, the analytics team updates the tracking code, changing how "conversion" is measured. The metric name stays the same. The dashboard updates. Nobody notices that the trend line now reflects a different calculation. Three months later, the marketing team reports a "20% improvement" that is actually a measurement change. A/B tests planned around the old definition no longer produce comparable results.

How Dawiso Supports KPI Management

Dawiso documents KPIs; it does not calculate them. Its business glossary holds business indicators alongside business terms and rules. Each indicator carries a description and a calculation, an owner and a steward, and can move through an approval workflow from draft to review to approved. That gives the "which revenue number is right?" debate one approved definition to settle on.

Indicators can be related to the reports and tables that compute them, and data lineage shows how those objects connect back toward the source system. Dawiso builds lineage from supported connectors, dbt, SQL parsing (not on by default) and manually added relations, and the depth depends on the source: Power BI lineage, for example, links reports, datasets and data sources at object level. When the churn number looks off, an analyst can follow that path to see which datasets and transformations feed the report.

Dawiso's data catalog lets teams discover which KPIs already exist before inventing new ones. Before a product team creates a new "engagement score," they can search the catalog for similar indicators and reuse the governed one instead of adding another conflicting definition.

Through the Model Context Protocol (MCP), AI agents can search Dawiso and read object details and lineage with the Dawiso MCP server. An analytics assistant asked "what is our churn rate?" can look up the governed indicator and its calculation before it queries data elsewhere, rather than guessing which table to use.

Conclusion

The hard part of KPIs is not selecting them. It is governing them - ensuring every team, every dashboard, and every automated report uses the same definition, the same source, and the same calculation. Organizations that treat KPI management as a metadata problem - with a governed glossary, lineage, and change management - get metrics they can trust. Those that skip governance get three different answers to the same question.

Sources

• Robert S. Kaplan and David P. Norton, The Balanced Scorecard - Measures That Drive Performance, Harvard Business Review, January-February 1992.
• George T. Doran, "There's a S.M.A.R.T. way to write management's goals and objectives", Management Review 70(11), November 1981, pp. 35-36.
• What Matters, OKR meaning, definition and examples.
• Google re:Work, Set goals with OKRs.
• Marilyn Strathern, "'Improving ratings': audit in the British University system", European Review 5(3), 1997, pp. 305-321.
• Dawiso Help, Business glossary.
• Dawiso Help, Business glossary attributes and components.
• Dawiso Help, Workflows overview.
• Dawiso Help, Sources of data lineage.
• Dawiso Help, Power BI connector.
• Dawiso Help, Dawiso Model Context Protocol (MCP).