You now have a test card, a matching destination, and a partner brief. The last piece of Class 72 is literacy: reading the numbers those efforts produce without panic and without self-deception. Ad metrics look like alphabet soup only until each abbreviation is tied to one question. This lesson ties each metric to its question, works a hypothetical example with visible arithmetic, and leaves you with a sheet that tells you what to fix next.

Each metric answers one question

Learn eight terms as eight questions:

• CPM (cost per mille): what did 1,000 impressions cost? Answers "how expensive is this attention?" Formula: spend ÷ impressions × 1,000.
• CPC (cost per click): what did one arriving visitor cost? Answers "how expensive is the handoff?" Formula: spend ÷ clicks.
• CTR (click-through rate): what share of impressions turned into clicks? Answers "did the message earn attention?" Formula: clicks ÷ impressions.
• CVR (conversion rate): what share of arrivals completed the action? Answers "did the destination and offer work?" Formula: conversions ÷ clicks.
• CAC (customer acquisition cost): what did one converted customer cost, all-in? Answers "can the economics survive?" Formula: total spend ÷ customers.
• ROAS (return on ad spend): how much revenue came back per unit of spend? Answers "did the money return?" Formula: attributed revenue ÷ spend.
• Frequency: how many times did the average person see the ad? Answers "are we informing or nagging?" Formula: impressions ÷ reach.
• Attribution window: how long after the click do we credit the ad for the action? Answers "which results belong to this test?" A 7-day click window credits conversions within seven days of the click; a 1-day view window credits conversions shortly after a mere view.

Two supporting ideas prevent misreading. Attribution is the rule assigning credit when several touches precede a purchase — the ad, the newsletter, the referral. No rule is neutral; pick one, state it, and compare tests only under the same rule. Incrementality asks the harder question: how many of these conversions would have happened anyway? A campaign can show a tidy ROAS and add little if it mostly harvests people who had already decided.

A worked hypothetical with honest arithmetic

This example is fictional and teaches logic, not a promise about any real campaign. A tailoring shop runs a two-week test: spend €200, reach 8,000 people, 20,000 impressions, 400 clicks, 20 bookings, 20 paying customers with an average order value of €35.

• CPM = 200 ÷ 20,000 × 1,000 = €10. Reaching a thousand screens cost ten euros.
• CTR = 400 ÷ 20,000 = 2%. Two in a hundred impressions earned a click — the message works adequately for local intent.
• CPC = 200 ÷ 400 = €0.50. Each arrival cost fifty cents.
• CVR = 20 ÷ 400 = 5%. One arrival in twenty booked — the destination converts acceptably.
• CAC = 200 ÷ 20 = €10 per customer.
• Revenue = 20 × 35 = €700. ROAS = 700 ÷ 200 = 3.5. Each euro returned three and a half in revenue — before garment costs, labor, and time, which ROAS never includes.
• Frequency = 20,000 ÷ 8,000 = 2.5. The average person saw the ad two and a half times: enough for recall, not yet nagging.

Now read it as decisions, not decoration. Suppose instead the CPC were a delightful €0.10 but CVR were 0.2%: cheap clicks, almost nobody booking. The cheap click is expensive — each customer would cost €50. The fix is the landing page or the offer, never more budget. Conversely, a frightening €40 CPM to 2,000 verified analysts can be viable for Sonariq if even a few start paid research plans: high reach cost, high customer value, sane CAC. Judge every metric against the value of the converted action, not against generic benchmarks quoted without context.

Frequency and attribution deserve their own glance. Frequency climbing past five or six with falling CTR often signals fatigue: the same people seeing the same creative stop responding. The answer is fresh creative or a wider audience, not higher bids. And if the shop counts walk-ins "from the ad" three weeks later under a 28-day view window, ROAS inflates with conversions the ad may not have caused. State the window on the sheet and keep it constant between comparisons.

Exercise: fill the metric sheet

Create AD-METRIC-SHEET.md with four blocks:

• Inputs: spend, impressions, reach, clicks, conversions, customers, average order value, attribution window.
• Formulas and results: CPM, CPC, CTR, CVR, CAC, ROAS, frequency — each with the numbers substituted, so anyone can audit the arithmetic.
• Interpretation: one sentence per metric answering its question ("CTR 0.4% — the message is not earning attention in this placement").
• Next action for three outcomes: (a) attention wins but action lags — inspect destination and offer; (b) action is strong but attention is thin — test new audiences or creative hooks; (c) everything is weak — pause, confirm tracking works, then change one piece.

Work all three outcome branches even if your current numbers favor one. The point of the sheet is a pre-committed decision rule, written before optimism or disappointment arrives. Note the incrementality caveat in one line: which conversions might have happened without the ad, and what would test that (a holdout area, a paused week, a partner-code comparison)?

Finish line: an AD-METRIC-SHEET.md with auditable inputs, computed metrics, a plain-language interpretation, and the next action for three possible outcomes.

Verify quickly: cover the results column and recompute one metric by hand from inputs. If the formula is not written clearly enough to redo, rewrite it.

Common failure mode: optimizing the cheapest number — celebrating low CPC or high CTR while CAC quietly exceeds what the product can bear. Always carry the chain through to the converted customer and the economics behind them.

Check your understanding

1. Which question do CPM, CTR, CVR, and CAC each answer, and which campaign piece does each indict when weak? 2. Recompute ROAS for the hypothetical if average order value falls to €20. What changes about the decision? 3. Why must two campaigns share an attribution window before their ROAS figures can be compared?

Next

Class 72 gave you the paid engine and its instruments. Class 73 adds the creative factory: Lesson 73.1 maps how AI accelerates the production line from brief to launch while human ownership stays where it matters.