Choosing the Right KPIs

This pack gives you 16 KPIs per dashboard — but you should report far fewer. The

hardest part of measurement isn't calculating metrics; it's choosing which ones

deserve a leader's attention. This guide helps you cut from 16 to the 5–7 that

actually drive decisions.

The problem with "track everything"

More metrics feel safer. They aren't. A dashboard with 40 numbers has the same effect

as a dashboard with zero: nobody knows where to look, so nobody looks. Every KPI you

add dilutes the ones that matter and adds maintenance cost. The goal is the

smallest set that lets you steer the business.

Rule of thumb: a team should be able to name its KPIs from memory. If they can't,

there are too many.

The test: is it a real KPI?

A metric earns a spot on the dashboard only if it passes all four:

1. Actionable — if it moves the wrong way, you know roughly what to *do*. A number

you can only watch is a *metric*, not a KPI.

2. Tied to an outcome — it connects to revenue, retention, cost, or risk. "Page

views" rarely does; "activation rate" does.

3. Owned — exactly one person or team is accountable for it (the Owner column).

A KPI everyone owns is a KPI no one owns.

4. Trustworthy — you can measure it consistently and people believe the number. A

precise metric from a flaky source is worse than a rough one from a solid source.

If a metric fails any of these, demote it to a secondary "diagnostic" you check only

when a real KPI moves.

Leading vs. lagging indicators

Balance the two or you'll fly blind:

A healthy dashboard pairs each lagging outcome with the leading metric that predicts

it. On the Sales dashboard, *New Bookings* (lagging) is predicted by *Pipeline

Coverage* and *SQOs* (leading). On Product, *Day-30 Retention* (lagging) is

predicted by *Activation Rate* and *Time to Value* (leading). Watch the leading ones

to change the lagging ones.

The "North Star" approach

Pick one metric that best captures the value your team delivers — your North Star —

then surround it with 4–6 supporting KPIs that explain its movement.

The North Star is the headline; the supporting KPIs are the diagnostics that tell you

*why* it moved.

How many is right?

When in doubt, fewer. You can always drill into a diagnostic when a KPI flags.

A worked example: cutting the Sales dashboard to 6

The Sales dashboard ships with 16 KPIs. For a weekly sales standup, keep:

1. Quota Attainment — the North Star; are we hitting the number?

2. Pipeline Coverage — leading indicator; do we have enough to hit next quarter?

3. Win Rate — efficiency; are we getting better at closing?

4. Sales Cycle Length — leading; are deals speeding up or stalling?

5. Average Contract Value — are we moving up-market or down?

6. Lead Response Time — the one activity metric that reliably predicts win rate.

The other ten (Forecast Accuracy, NRR, Upsell Rate, CAC, Revenue per Rep…) are

valuable diagnostics — review them monthly or when a top-6 KPI breaks, not in the

weekly glance.

Common KPI mistakes

• Vanity metrics: big numbers that only go up and don't inform a decision (total

registered users, lifetime page views). Impressive, useless.

• Proxy obsession: optimizing the measurable stand-in until it stops reflecting the

goal (Goodhart's Law — "when a measure becomes a target, it ceases to be a good

measure"). Watch the outcome, not just the proxy.

• Unowned metrics: a KPI with no name next to it never improves.
• No target: a number without a target is trivia. Set one (see

docs/CUSTOMIZATION.md §2), even if it's a rough first guess.

• Mixing directions silently: half your KPIs are "higher is better" and half

"lower is better." The Direction column (formulas §0) keeps the scoring honest —

use it.

Your selection checklist

For each KPI you're considering, ask:

• [ ] If this moves the wrong way, do I know what to do?
• [ ] Does it connect to revenue, retention, cost, or risk?
• [ ] Is there exactly one owner?
• [ ] Do I trust the number enough to act on it?
• [ ] Is it leading (steer) or lagging (score) — and do I have a balance of both?
• [ ] Could I remove it and lose nothing? (If yes — remove it.)

Get this right and guides/dashboard-design-principles.md will help you present the

survivors so people actually use them.

Dashboard Design Principles

A dashboard's job is to drive a decision in seconds. Most fail not because the data is

wrong, but because the *presentation* makes the reader work too hard. This guide

covers the design choices that turn a wall of numbers into something a busy leader

actually reads — and how this pack's layout already bakes them in.

The five-second test

Show your dashboard to someone for five seconds, then hide it and ask: *"What needs

attention?"* If they can't answer, the design has failed. Everything below serves that

test.

1. Lead with status, not data

The eye should land on what's wrong before any raw numbers. That's why the pack

puts a Status traffic light (column H) on every row and a vs-target heatmap on

column G. A reader scans the red and yellow first, ignores the green, and drills in

only where needed.

• Green = on or above target — no action.
• Yellow = within 5% under target — watch it.
• Red = more than 5% under — act now.

Apply the conditional formatting from formulas/FORMULAS.md §5 and the dashboard

becomes a scan, not a read.

2. Always show context, never a number alone

A bare number is meaningless. Win Rate: 27% — good or bad? The pack pairs every

Current_Value with three pieces of context:

1. A target (column E) — is 27% where we wanted to be?

2. A comparison (MoM_Change_Pct, column F) — up from 24% or down from 30%?

3. A trend (the sparkline) — a steady climb or a one-off spike?

*Number + target + direction + trend* is the minimum viable KPI display. Each alone

misleads; together they tell the truth.

3. Make the trend visible (sparklines earn their space)

A single sparkline communicates more than a column of monthly figures. It shows

direction, volatility, and recent inflections at a glance. The pack provides a 12-month

Trend_12mo series on every row precisely so you can drop in a sparkline (formulas

§4). A green status with a downward sparkline is an early warning a static number

would hide — that's the whole point.

4. Respect the visual hierarchy

Guide the eye deliberately:

• Freeze the header row (View → Freeze → 1 row) so labels never scroll away.
• Group by category (column B) — cluster related KPIs so the reader builds a mental

map. Use QUERY (formulas §6) to produce category-grouped views.

• Right-align numbers, left-align text. Aligned decimal points are scannable;

ragged ones aren't.

• Put the North Star first. The metric that matters most goes at the top, not

buried in alphabetical order.

• Use whitespace. A thin separator row between categories does more for readability

than any color.

5. Don't rely on color alone

Roughly 1 in 12 men has some color vision deficiency. A red/green-only dashboard is

illegible to them. Mitigations the pack supports:

• Keep the text labels Green/Yellow/Red in column H — the word carries the

meaning even if the color doesn't.

• Use shape or position as a second channel: an up/down arrow on the delta

(formulas §5.3), or sorting worst-to-best so position encodes severity.

• Choose a colorblind-safe palette — blue/orange reads more reliably than

red/green if you can't keep the labels.

6. Round aggressively

Precision implies confidence you don't have. Conversion Rate: 3.7% is honest;

3.6938% is false precision that slows reading and invites pointless debate about the

fourth decimal. Match the displayed precision to the decision it informs:

• Big-picture KPIs: whole numbers or one decimal.
• Money: nearest dollar (or thousand, for large figures) — use $285K, not

$285,142.61.

• Percentages: one decimal is almost always enough.

Format for display; keep full precision in the underlying cell.

7. Design for the audience and cadence

The same data needs different presentations:

Don't show executives 160 numbers and don't show the team a single rolled-up score.

Build the one-page Scorecard (formulas §7) for leadership and keep the full

dashboards for the people who act on them.

8. Annotate the "why"

A number tells you *what*; a one-line note tells you *why* — and *why* is what prevents

panic and bad decisions. When a KPI moves sharply, add a comment or a Notes cell:

"Churn spike = one enterprise logo offboarded, not a trend." Without it, every reader

invents their own (usually wrong) explanation. The pack's Owner column tells readers

*who* to ask; a notes habit tells them the answer before they have to.

9. Keep it current or kill it

A stale dashboard is worse than none — it erodes trust in every number on it. Set a

realistic cadence per KPI (the Update_Frequency column), automate what you can

(docs/CUSTOMIZATION.md §7), and if a dashboard consistently goes stale, that's a

signal nobody's using it — cut it.

The design checklist

Before you share a dashboard, confirm:

• [ ] A reader can answer "what needs attention?" in five seconds.
• [ ] Status (traffic light) is the first thing the eye lands on.
• [ ] Every number has a target, a comparison, and a trend.
• [ ] Sparklines are present and readable.
• [ ] Header row frozen; KPIs grouped by category; North Star on top.
• [ ] Meaning survives in greyscale (labels/arrows, not color alone).
• [ ] Numbers are rounded to decision-relevant precision.
• [ ] The view matches the audience (team vs. leader vs. exec).
• [ ] Sharp moves are annotated with a one-line "why."
• [ ] The data is current.

Nail these and the dashboards in this pack stop being spreadsheets and start being

decision tools. Pair this with guides/choosing-the-right-kpis.md to make sure you're

presenting the *right* metrics beautifully — not the wrong ones.