All about KPIs: numbers that drive a decision

The weekly meeting opens with a slide of 28 charts. Someone says “we’re green on KPIs.” Two minutes later the room argues about a dip nobody can explain. Three metrics move because of a tracking change. One “KPI” is just page views with a fancy name. You leave with more screenshots than decisions.

A key performance indicator (KPI) is a small set of measures tied to a goal and a decision. Everything else is a metric, a diagnostic, or noise. This post helps you choose KPIs that change behavior, avoid vanity traps, and build a simple operating rhythm. It pairs well with reading a number like an adult and the one-page analytics brief.

What you’ll learn

KPI vs metric vs target vs OKR (without consulting fog)
How to pick a short list with owners and decision rules
Leading vs lagging indicators in practical terms
A ladder from activity noise up to decisions
Failure modes: too many KPIs, set-and-forget, bad quality

Definitions that fit on a sticky note

Term	Meaning	Example
Metric	Any measured quantity	Tickets opened per day
KPI	A metric that steers a goal	Median first-response time
Target	The number you aim for	≤ 4 minutes
Diagnostic	Explains movement in a KPI	Staffing hours, ticket mix
OKR	Goal framework; KPIs often measure KRs	Objective: faster support; KR: p50 response ≤ 4m

If nothing would change in your calendar when the number moves, it is probably not a KPI. It might still be useful. Just do not put it on the executive strip.

The metric ladder

Climb the ladder before you print the slide.

Work top-down. Start from the decision (“hire agents if wait is bad for two weeks”), then the outcome KPI, then diagnostics. Bottom-up “we track everything” creates dashboard landfills.

How to choose KPIs

1. Start from a decision and a goal

Write the decision in a sentence. “If trial-to-paid conversion stays under 8% for two weeks after the pricing test, we roll back.” Then pick the measure that would trigger that decision. If you cannot name the decision, you are decorating.

2. Prefer fewer

A team can truly steward a handful of KPIs. Beyond that, attention dies. Many teams run 3-5 team KPIs and keep a deeper diagnostic pack for analysts.

3. Make the definition boring and exact

Grain, filters, time zone, and owner. Same discipline as governance. “Conversion rate” without a denominator story is a fight waiting to happen.

KPI: trial_to_paid_7d
Numerator: accounts with first paid invoice within 7 days of trial start
Denominator: trials started in period (exclude internal + QA accounts)
Time zone: UTC
Owner: Growth analytics
Decision rule: < 8% for 14 days post-change → recommend rollback

4. Balance leading and lagging

Type	What it tells you	Example
Lagging	Outcomes after the fact	Monthly recurring revenue, churned accounts
Leading	Earlier signals you can still act on	Activation rate in first 24h, sales cycle stage movement

Lagging without leading means you learn late. Leading without lagging means you optimize theater. Use both, with clear links.

5. Assign an owner and a review rhythm

Owner means “person who explains movement and drives the response,” not “person who built the chart once.” Weekly for operational KPIs; monthly for slower ones is a common pattern. Reviews without owners are slideshows.

Worked example: support team strip

KPI	Why it is key	Diagnostic friends
Median first-response time	Customers feel speed here	Staffing, volume by channel
% resolved in one touch	Quality of answers	Macro usage, reopen rate
CSAT on closed tickets	Perceived quality	Topic tags, agent cohort

Notice what is not on the strip: raw ticket count alone. Volume matters for staffing models, but celebrating “more tickets” as success is how you reward chaos.

Good habits

Show the definition next to the chart (tooltip or linked card)
Annotate tracking changes on the timeline
Compare to a baseline period and, when relevant, a target band
Separate scorecard KPIs from sandbox exploration
Retire KPIs that no longer map to a decision

Things to watch out for

Too many KPIs

If everything is key, nothing is. Cut until each remaining KPI has an owner and a decision rule.

Vanity metrics

Big numbers that always go up (raw page views, emails sent) feel comforting. Ask what decision they change. If the answer is “awareness,” dig one level deeper into quality of awareness.

Set and forget

Business mix shifts. A KPI that mattered at 20 employees can mislead at 200. Schedule reviews of the KPI list itself, not only the values.

Quality issues

A beautiful KPI on broken events is fiction. Pair KPIs with basic quality checks (null rates, duplicate events, timezone bugs). When tracking breaks, mark the chart instead of narrating a fake story.

Misaligned incentives

If you pay people on a metric they can game, they will game it. Pair quantity with quality, or audit samples. Goodhart’s law is not a joke; it is a planning assumption.

KPIs and strategy documents

OKRs, scorecards, and north-star frameworks can all work. The failure mode is the same: copying a template metric list from another company. Your KPI set should fall out of your strategy and operating decisions, not from a generic SaaS blog. Steal structure, not numbers.

A lightweight weekly ritual

Open the strip (3-5 KPIs) with definitions visible
For each mover: what changed, is it real, what will we do?
Log one action or explicit “watch only”
Park deep dives in a separate analysis ticket with a brief

Thirty minutes. Ruthless parking lot. That beats a two-hour tour of 40 tiles.

Quick recap

KPIs are few, owned, and tied to decisions; metrics can be many
Write exact definitions, targets, and decision rules
Use a ladder: decision → outcome KPI → diagnostics → activity
Mix leading and lagging; watch for vanity and gaming
Review the list itself on a calendar, not only the sparkline

When two KPIs conflict (speed vs quality), do not hide the tradeoff in a weighted score nobody understands. Show both. Decide the policy in the open (“we will not let CSAT fall more than 3 points to hit response time”). Composite indices can help later; they are a poor first move when trust is low.

For leadership audiences, pair each KPI with a single plain sentence on what good looks like this quarter. Numbers without narrative get misread. Narrative without numbers is a pep talk. You want both, short.

If you are rebuilding a scorecard after a reorg, archive the old one with a date stamp instead of silently editing history. People remember screenshots. A clear “scorecard v3 effective June 1” note saves conspiracy theories about moved goalposts.

Write the messy edge cases in the open. Hidden footnotes become tribal knowledge and then become outages.

If two teams need different definitions, name both clearly instead of forcing a fake compromise that satisfies nobody.

Ship the smallest useful artifact this week: a definition card, a quality check, or a retired vanity chart. Momentum beats manifesto.

Teach newcomers where the source of truth lives. Onboarding is a governance surface whether you designed it or not.

When something fails, prefer a short postmortem over a new committee. Fix the rule or the test that should have caught it.