Module 6: Reading and Improving

CTR, CPC, CPA, and ROAS in plain English

Four metrics run the whole game. CTR (click-through rate) is clicks divided by impressions — it measures whether your creative earns attention. Around 1 percent or better is decent on Meta's feed; below 0.5 percent usually means the creative is the problem. CPC (cost per click) measures how expensive your auction is. CPA (cost per acquisition) is spend divided by conversions — the number that decides whether your ads work as a business. ROAS (return on ad spend) is revenue divided by spend — 3.0 means $3 back for every $1 spent.

Kill, keep, or scale decisions

Set a weekly review ritual — same day, same time, ad set level, last 7 days of data. Decisions come in three flavors. Kill: the ad set has spent two to three times your target CPA with zero conversions — cut it, no sentimentality. Also kill individual creatives with no engagement after $20 to $30 of spend (CTR under 0.5 percent, costs climbing). Keep: conversions are coming in at or near your target CPA — leave it completely alone. Touching a working ad set to "optimize" it is how winners get broken.

Scale is the fun one, and it has rules. Only scale what has beaten your target CPA for at least seven straight days — not two good days. When you scale, raise the budget 20 to 30 percent, wait a few days, and repeat. Doubling overnight usually collapses performance because it throws the ad set back into learning with a shocked system. The alternative scaling move: duplicate the winning ad set into a new campaign at the higher budget and let the original keep running — if the duplicate flops, you still have the winner intact. Never scale and edit creative at the same time; if performance changes, you won't know which move caused it.

Keep a simple decision log — a notebook or spreadsheet with date, what you changed, and why. "Oct 21: killed interest ad set, $68 spent, 0 leads, target CPA $30." This log becomes your most valuable asset within two months, because it stops you from re-testing things you already learned. The emotional pitfalls: killing a day-2 ad set that's still in the learning phase, keeping a loser alive out of hope ("maybe it'll turn around" — it won't, cut it), and checking the dashboard six times a day. Weekly decisions, made calmly, beat hourly anxiety every time.

Testing one thing at a time (methodical iteration)

If you change the hook, the image, and the audience all at once and performance improves, you've learned nothing — you don't know which change did it, so you can't repeat it. Real testing changes one variable at a time: one new hook against the same image and audience, or one new image against the same hook and audience. One change, one result, one lesson. It's slower than guessing and roughly ten times more effective.

Run every test as a loop: write down your hypothesis ("a testimonial hook will beat the pricing hook for cold traffic"), make the single change, run it for seven days or until it has meaningful spend behind it, read the result, log it, and move to the next test. Test in this order of leverage: hooks and creative first (the biggest lever), then the offer, then the audience, then the landing page. Keep the log from the last lesson going — date, what changed, spend, result. Over months, this log compounds into an unfair advantage: you stop paying to re-learn the same lessons.