luvv to helpDiscover the Best Free Online Tools
Topic 7 of 12

Metric Definitions and Glossary

Learn Metric Definitions and Glossary for free with explanations, exercises, and a quick test (for Data Analyst).

Published: December 20, 2025 | Updated: December 20, 2025

Table of Contents

Why this matters

Clear metric definitions are the backbone of trustworthy BI dashboards. As a Data Analyst, you will:

  • Translate business questions into precise, repeatable metrics.
  • Prevent disputes by documenting inclusions, exclusions, and time grain.
  • Enable consistent decisions across teams and tools.
  • Speed up onboarding with a shared glossary everyone can use.
Real tasks you will face
  • Audit conflicting KPIs across teams and create a single definition.
  • Write a metric card that product, marketing, and finance all sign off on.
  • Explain why a dashboard number changed after a data model update.
  • Add a new KPI while preserving historical comparability.

Concept explained simply

A metric definition is a recipe for a number. If two people follow the same recipe, they should get the same result.

Mental model: the Metric Card

Think of a Metric Card as a small spec that answers 10 key questions. If any answer is missing, the number can drift.

Metric Card — the 10 must-haves
  1. Name and short purpose
  2. Business question it answers
  3. Formula with explicit numerator and denominator
  4. Time grain (daily/weekly/monthly) and window (e.g., rolling 7 days)
  5. Dimensions allowed for slicing (country, channel, etc.)
  6. Filters and exclusions (bots, test traffic, canceled orders)
  7. Data source tables/views and join keys
  8. Inclusion logic (unique users vs sessions, net vs gross)
  9. Refresh cadence and data freshness expectations
  10. Owner and QA checks, plus an example calculation

Anatomy of a good metric definition

  1. State the business question. Example: Are we turning website traffic into purchases?
  2. Write the formula plainly. Use words first, then the precise calculation.
  3. Lock the time grain. Daily vs weekly can change values.
  4. Clarify populations. Users vs sessions; paying customers vs signups.
  5. List all filters. Remove test data, internal staff, and refunded orders if needed.
  6. Declare dimensions. Which segment splits are valid?
  7. Name the sources. Tables/views and their keys.
  8. Define refresh & quality checks. When does it update and how do you validate it?
  9. Give a worked example. Show numbers from a small mock dataset.
  10. Assign an owner. Who approves changes?

Worked examples

Example 1 — E‑commerce Conversion Rate

Name: Site Conversion Rate (Daily)
Question: What share of unique visitors complete a purchase today?
Formula (words): Purchases today divided by unique visitors today.
Formula (precise): count_distinct(orders.order_id where order_status = 'paid' and order_date = d) / count_distinct(web_visits.user_id where visit_date = d)
Grain: Daily (UTC), no rolling window.
Filters: Exclude staff/test users and orders with full refunds.
Dimensions: traffic_source, device_type, country.
Sources: orders, web_visits (join on user_id not required here).
Refresh: Hourly; finalized EOD.
Owner: Growth Analytics.
Example:

Day d: unique visitors = 1,000
Paid orders (net) = 55
Conversion Rate = 55 / 1,000 = 5.5%

Edge cases: If visitors = 0, return null and flag data quality.

Example 2 — Subscription Monthly Churn Rate

Name: Subscriber Churn Rate (Monthly)
Question: What percent of paying subscribers leave this month?
Formula (words): Subscribers who churned this month divided by subscribers at start of month.
Formula (precise): churned_subs_m / opening_subs_m
Grain: Monthly, calendar months.
Filters: Exclude involuntary churns reversed within 7 days (billing retries).
Dimensions: plan_tier, country.
Sources: subscriptions_snapshot_monthly.
Refresh: Monthly +3 days for finalization.
Owner: Revenue Analytics.
Example:

Opening subs on Mar 1 = 20,000
Churned during Mar (net of reactivations within 7 days) = 1,100
Churn Rate = 1,100 / 20,000 = 5.5%

Edge cases: If opening_subs_m = 0, return null and alert.

Example 3 — Average Order Value (AOV)

Name: AOV (Daily)
Question: How much revenue do we get per order today?
Formula (words): Net revenue divided by number of paid orders.
Formula (precise): sum(orders.net_revenue where order_status='paid') / count_distinct(order_id where order_status='paid')
Grain: Daily.
Filters: Exclude shipping revenue if policy defines net_revenue without shipping; exclude full refunds.
Dimensions: traffic_source, device.
Sources: orders.
Refresh: Hourly.
Owner: Commerce Analytics.
Example:

Paid orders = 80
Net revenue = $6,800
AOV = 6,800 / 80 = $85.00

Write your glossary

Use this template for each metric.

Copy-friendly Metric Card template
Name:
Purpose (1 sentence):
Business question:
Formula (words):
Formula (precise):
Numerator:
Denominator:
Time grain and window:
Valid dimensions:
Filters and exclusions:
Inclusions logic (unique users? net vs gross?):
Data sources and keys:
Refresh cadence and freshness expectation:
Owner and change process:
QA checks:
Example calculation (with small numbers):
Notes (edge cases, known gaps):

Exercises

Everyone can do the exercises and take the Quick Test. Only logged‑in users will have their progress saved.

Exercise 1 — Define WAU precisely

Create a Metric Card for Weekly Active Users (WAU).

  • Business context: A user is active if they trigger the event app_open or completes any meaningful action (purchase, message_sent, file_upload).
  • Time grain: Weekly (Mon–Sun), reported on week ending Sunday.
  • Requirement: Unique users per week; exclude staff/test accounts.

Deliverable: A filled Metric Card with an example using a tiny mock week.

Exercise 2 — Calculate three metrics from mock data

Using the dataset below, compute Conversion Rate, AOV, and Monthly Churn Rate.

Visitors (April 10): 1,200 unique
Orders (April 10): 66 paid, 4 fully refunded
Order net revenue (after refunds) for April 10: $5,940
Subscribers opening on April 1: 12,000
Churned in April (net of 2 reactivations within 7 days): 720
  • Conversion Rate uses paid orders net of full refunds as numerator.
  • AOV uses net revenue and count of paid orders (exclude fully refunded orders).
  • Churn Rate uses net churn divided by opening subs.

Checklist before you submit

  • Time grain stated and unambiguous.
  • Numerator and denominator clearly defined.
  • All exclusions listed (staff, test, refunds).
  • Sources and refresh cadence included.
  • One small example with numbers.

Common mistakes and self-check

  • Undefined population: Mixing users and sessions. Self-check: Does the denominator say unique users or sessions?
  • Time grain drift: Weekly metric calculated with daily logic. Self-check: Is the window and week boundary specified?
  • Missing exclusions: Internal traffic inflating KPIs. Self-check: Are test/staff IDs filtered?
  • Gross vs net confusion: Not handling refunds. Self-check: Does definition say net of refunds?
  • Implicit joins: Ambiguous source tables. Self-check: Are tables and keys listed?
  • Divide-by-zero handling: Returning 0 instead of null. Self-check: What happens when denominator is 0?
Quick self-audit steps
  1. Read the metric aloud to a non-analyst; if they can restate it, it’s clear.
  2. Have a peer compute the example independently; results must match.
  3. Test with a zero/edge case to verify your guards.

Practical projects

  • Create a 1‑page glossary for 8–10 core KPIs (traffic, conversion, revenue, retention). Acceptance: Every KPI has formula, grain, filters, owner, example.
  • Instrument a data quality checklist for two metrics (freshness threshold, row counts, zero‑denominator alert). Acceptance: Alerts trigger on anomalies in a sample extract.
  • Run a metric alignment workshop with a mock team. Acceptance: Resolve one conflict (e.g., net vs gross revenue) and document the final definition.

Learning path

  1. Review business goals and choose 5–10 essential KPIs.
  2. Draft Metric Cards using the template.
  3. Validate with small mock datasets and peer review.
  4. Implement calculations in your BI tool with identical logic.
  5. Set up refresh cadence and basic QA checks.
  6. Version your glossary; track and communicate any definition changes.

Mini challenge

Your PM asks for “activation rate.” Draft a concise first-pass definition:

  • Business question and formula (words + precise).
  • Time grain, population, and one example.
  • List at least two exclusions.
Hint

Activation often means a user completes a key action within N days of signup. Define N, the action, and unique user logic.

Next steps

  • Convert two of your Metric Cards into implemented measures in your BI tool.
  • Share your glossary draft for cross‑functional sign‑off.
  • Take the Quick Test below to confirm understanding.

Practice Exercises

2 exercises to complete

Instructions

Create a complete Metric Card for WAU (Weekly Active Users).

  • Active if user triggers app_open or any meaningful action (purchase, message_sent, file_upload).
  • Time grain: Weekly (Mon–Sun), reported on week ending Sunday.
  • Use unique users per week; exclude staff/test accounts.
  • Include data sources, keys, refresh cadence, QA checks, and one numeric example.
Expected Output
A finished Metric Card covering: Name, Purpose, Business question, Formula (words + precise), Numerator, Denominator, Time grain/window, Valid dimensions, Filters/exclusions, Inclusion logic, Data sources/keys, Refresh cadence, Owner, QA checks, Example calculation, Notes.

Metric Definitions and Glossary — Quick Test

Test your knowledge with 8 questions. Pass with 70% or higher.

8 questions70% to pass
Defining KPIsData Modeling Basics

Have questions about Metric Definitions and Glossary?

AI Assistant

Ask questions about this tool