KPI Fundamentals
Metric + target + owner + review
A KPI is a metric plus a target and a decision: outcome-oriented, measurable, owned, reviewed on a cadence. KPIs work because they compress a decision into a number and a delta.
What you will learn
- Define what makes a metric a KPI (vs a metric)
- Pair KPIs with decisions and owners
- Recognize vanity metrics and KPI anti-patterns
What is this?
A KPI is a metric that a decision hangs on: measurable, tied to a target, owned by someone accountable, and reviewed on a fixed cadence. The KPI design art is choosing the few that actually drive behavior — trading vanity metrics (great to report, useless to decide) for outcome metrics with a clear action when they drift.
Why do you need it?
- Most 'KPI dashboards' are metric walls — no target, no owner, no decision.
- The right 5 KPIs beat 40 chart-filled pages.
- Target + delta is what turns a number into a trigger to act.
- Designing KPIs is a core BI-analyst interview skill.
Mental model
A car's gauges you're actually using
KPIs
A car has dozens of gauges; you scan three: speed, fuel, temperature — each with a zone that says 'act now'. A mass-of-dials dashboard is the metric wall. KPIs are the dials tied to a decision ('refuel', 'slow down') with an owner behind the wheel.
How it works
- Pick the outcome the decision needs: 'revenue' not 'page views'.
- Define it precisely (formula, grain, caveats) once, like a model metric.
- Bind a target and a review cadence (daily/weekly/monthly).
- Assign one owner accountable for movement and interpretation.
- Display as value + delta vs target (green/red), not a giant chart wall.
From metric to decision-ready KPI.
Examples
Metric vs KPI — the test
PAGE VIEWS -> metric (vanity: no decision)MRR + % to target -> KPI (owner: revenue, monthly review)NPS score TTM -> metric (lagging, reported)Active users / cohort -> KPI (action: this month's cohort is off) KPI test: does this number, with its target,change a decision this period? YES -> KPI (few, owned, reviewed) NO -> metric (report background, not a dashboard centerpiece)Value-delta-target-owner-machine: the KPI test filters the metric wall into decision triggers.
In production
- An exec board runs 5 KPIs with targets and owners — decision minutes, not data dumps.
- Support tracks SLA-fill and CSAT as KPIs; each has a threshold that pages the owner.
- Growth uses activation %, not raw registrations, as the owned KPI.
Best practices
- Fewer is stronger: 5-7 owned KPIs per leadership view, not 30.
- Make every KPI a precise model metric with an owner.
- Targets are set, reviewed and re-negotiated on the cadence.
- Show value, delta, and traffic-light vs target — the decision at a glance.
Common mistakes
- KPI walls with no target or owner.
- Vanity metrics ('sessions') wearing KPI clothes.
- KPIs that never change behavior (no cadence review).
- Owned by nobody — everyone quietly ignores the red light.
Debugging tips
- KPI red-light ignored → no owner/cadence review; fix the governance, not the chart.
- Dashboards 'look busy' but change nothing → run the KPI test and cut the wall.
- Targets drifting mid-year → target renegotiation happening anywhere; schedule it.
Interview checkpoint
Hands-on lab
- Write the KPI test as a checklist from real metrics
- Calculate value/delta/target vs a control
- Decide which metrics survive as KPIs
Score the KPI Test on Metrics You Design
40 min · beginner
Quick quiz
Which element turns a metric into a KPI?
Page views with no decision attached is a strong KPI.