The dashboard behind the dashboard

You can read a query and judge a claim. Now we point that skill at the artifact you see most: the dashboard. Every KPI tile - active subscribers, revenue, churn - is a one-line SQL query someone saved months ago and rarely revisits. Understanding that a tile is a query, not a truth, is what lets you ask why two "subscriber" counts disagree and which one belongs in the decision.

A KPI tile is the friendly face on a saved SELECT. The big number, the little label, the trend arrow - all of it is the result of a query the BI tool runs on a schedule. Peel it back and the definition choices become visible: which table, which filter, which grain. That is where the real meaning lives.

SELECT COUNT(*) AS active_subs
FROM subscriptions
WHERE cancel_date IS NULL;

The most common dashboard confusion is one word carrying two definitions. "Subscribers" can mean everyone who ever subscribed, or only those active right now - and a dashboard can show both on different tiles, each true. The trap is a denominator game: the word stays the same while the population underneath quietly changes.

SELECT COUNT(*) AS active_subs
FROM subscriptions
WHERE cancel_date IS NULL;

SELECT COUNT(*) AS all_subs
FROM subscriptions;

A rate is a fraction, and it is only honest when the top and the bottom describe the same population over the same window. Churn rate is the classic offender: churned customers divided by... which customers? All-time? Active at the start of the month? The denominator you choose can move a churn number by a wide margin without a single customer changing.

SELECT substr(signup_date, 1, 7) AS cohort,
       COUNT(*) AS new_customers
FROM customers
GROUP BY cohort
ORDER BY cohort;

Take one KPI from a dashboard you actually use. As a group, write its likely one-line query definition, then name the trap to check. You are not writing production SQL - you are stating the definition precisely enough that an analyst could confirm or correct it.

SELECT COUNT(*) AS active_subs
FROM subscriptions
WHERE cancel_date IS NULL;

SELECT COUNT(*) AS completed_orders
FROM orders
WHERE status = 'completed';

• Pick the single most important tile on your main dashboard. Write its likely one-line query definition, then send your analyst the one-sentence question that would confirm it. Compare their answer to your guess.
• Find any two tiles that use the same word - "users", "customers", "revenue". Check whether they share a definition. If they differ, you have found a boardroom argument waiting to happen; get ahead of it.
• For one rate on your dashboard, write down its numerator and denominator explicitly. Confirm they cover the same population and window. If they don't, flag it.
• Optional: preview session a5, where we turn all of this into a precise data request - asking an analyst for exactly the right cut, the first time.

This session connects SQL reading to the BI layer leaders live in every day - dashboards and KPIs - and the definition traps that hide inside them. It covers: