What makes a metric a metric
A metric is a number that:
- Can be computed, has an explicit formula or aggregation rule
- Means something to a stakeholder, answers a business question
- Can be compared, over time, across slices, against targets
Raw warehouse columns aren’t metrics. A column called orders.total_price_cents is data. “Revenue” is a metric defined as SUM(orders.total_price_cents) / 100 filtered to non-refunded orders.
Metric definition principles
The hard part isn’t computing metrics, it’s deciding what each one means. Three principles:
- One canonical definition per metric. Not three flavours, one. If the business genuinely needs three flavours of ROAS, give them three different names.
- Definitions live in code, not in heads. A spreadsheet of metric formulas drifts. A semantic-layer model in git is enforceable.
- Definitions change deliberately, with versioning. When you re-define “Revenue” to exclude refunds, every report’s “Revenue” series shifts on the change date. Document the cutover.
Metric vs KPI
Often used interchangeably. Subtly different:
- Metric, any computed business number
- KPI (Key Performance Indicator), a metric the business has elevated to “this is what we’re optimising against”
You have hundreds of metrics. You have maybe 5-10 KPIs. KPIs get board attention. Metrics get analyst attention.
Metric categories in marketing analytics
A useful taxonomy:
- Spend metrics, ad spend, by channel, by period
- Acquisition metrics, new customers, CAC, conversion rate, cost per lead
- Revenue metrics, revenue, AOV, ROAS, blended ROAS
- Retention metrics, LTV, repeat rate, churn, cohort revenue
- Efficiency metrics, payback period, contribution margin, ratio metrics
Strong reporting typically combines metrics from each category, pure spend metrics without efficiency context, or pure revenue metrics without spend context, miss the picture.
Metric pruning
A symptom of a struggling analytics function: 400+ metrics defined, most never queried. Each adds maintenance burden and definitional ambiguity.
Healthy practice: audit metric usage quarterly. If a metric hasn’t been queried in 90 days, archive it. If two metrics are near-duplicates, merge them.
Common mistakes
- Letting every team define their own metrics. Fragments definitions. Same name means different things in different rooms.
- Defining metrics in the BI tool, not the semantic layer. Couples the metric to the BI tool. Move to a tool-agnostic semantic layer.
- Confusing “more metrics” with “more insight.” They aren’t. Insight comes from understanding the few metrics that matter deeply.
FAQ about Metric
What is a metric?
A metric is a quantitative measurement of business performance, revenue, orders, ROAS, CAC. It has an explicit formula, means something to a stakeholder, and can be compared over time or across slices.
What is the difference between a metric and a KPI?
Every KPI is a metric, but only a few metrics are KPIs. A KPI is a metric the business has elevated to “this is what we are optimising against.” Most organisations have 3-10 KPIs and hundreds of metrics.
Where should metrics be defined?
In a semantic layer, in code, version-controlled in git. Defining metrics in BI tools or in individual dashboards leads to drift, the same “Revenue” ends up calculated differently in different reports.