Metrics and analysis
A weekly optimization routine: what to check, in what order, every Monday
One hour per account, always the same, in the order that stops you fixing the wrong thing. The script, the thresholds and three questions before any edit.
Also available in: Português · Español
Optimization without a routine becomes reaction: someone looks at the dashboard, sees an ugly number, changes something. Next week, another ugly number, another change. Three months later, nobody knows why the account looks the way it does.
This is a routine of about an hour per account, always in the same order, on the same day. The order matters more than the content.
The order, and why it is that one
1. Measurement is it counting correctly?
2. Result what changed this week?
3. Creative did a piece fatigue?
4. Audience did one exhaust?
5. Budget where to move money?
Most people start at 5 and end at 1 — if they get there. That path produces the most work and the least result, because budget is the most expensive thing to touch and the least frequent cause.
1. Measurement (5 minutes)
Before looking at any number, confirm the number exists:
- Did the conversion event fire in the last 7 days?
- Is event volume inside its normal range?
- Is any ad or ad set rejected?
A CPL that doubled with the event stopped is not a campaign problem. It is a tag problem, and no optimization fixes it.
Submit a test lead once a month. It takes two minutes and eliminates the most expensive cause on the list.
2. Result (15 minutes)
Compare this week against the previous three, never against one:
| what to look for | |
|---|---|
| Spend | did it vary more than planned? |
| Leads | did they fall with spend, or alone? |
| CPL | inside the account's normal range? |
| Lead → sale conversion | this is where the diagnosis splits |
The last row is what stops you touching the campaign when the problem is sales — the full test is in expensive leads or a broken funnel.
3. Creative (15 minutes)
Open the active ads and look for three things:
Concentration. One ad taking nearly all the spend means the others are not really being tested.
Fatigue. Frequency rising while CTR falls is the classic pattern; the mechanism is in creative fatigue.
What sold. If you have revenue by creative, this is the moment to look — and that ranking usually disagrees with the CPL ranking.
Rule: one new piece in rotation per week, before you need it. Swapping creative after CTR has fallen always costs more.
4. Audience (10 minutes)
Only enter here if steps 1 to 3 did not explain things:
- reach no longer growing
- frequency above the account's normal
- CPM rising in specific ad sets
If CPM rose across the whole account, it is not your audience — it is the auction, and the action is to wait.
5. Budget (10 minutes)
Last, and in steps:
Raise where cost per sale sits below target and there is volume. At most 20–25% at a time.
Cut where spend passed the threshold with no result. A threshold, not an impression.
Leave alone whatever sits inside the range. Changing for the sake of it restarts learning and costs money.
Three questions before any edit
Before clicking save, answer:
- Which cause did I confirm? If the answer is "I thought maybe", go back to the matching step.
- What signal will tell me it worked? Define it beforehand, or you will judge by the overall result, which has ten variables in it.
- When do I look again? Put a date on it. Changing again within 48 hours is the most common way to never learn what caused what.
If you cannot answer all three, it is not time to change anything.
The log
One line per change, always in the same place:
13 Oct · Ad set "SP-interests" paused
Cause: spent $1,900 over 14 days, 2 leads, 0 quotes
Expected signal: campaign CPL falling below $70
Review: 20 Oct
It looks like bureaucracy and it is what lets you know, three months later, why the account looks the way it does. Without it, the decision gets remade from zero every time someone new looks.
What does NOT belong in the weekly routine
Strategy changes. Offer, positioning and account structure are monthly or quarterly decisions. Mixing them into the weekly routine produces an account that never stabilizes.
A new creative every day. One a week, in rotation. More than that prevents knowing which one worked.
Pausing on the first bad day. Three days in the same direction is a trend; one day is noise.
Where to start
In the first week, optimize nothing. Just walk the five steps and write down your thresholds: what a normal CPL is, what frequency counts as high, how much spend without results you accept.
Without those numbers, every decision becomes a comparison against the market — and the market does not pay your bills.
For what to do when the reading points at sales, see good campaign, bad follow-up.
Frequently asked questions
How often should I optimize a campaign?
Weekly for decisions, daily only for anomalies. Touching it every day prevents knowing what caused what, because every meaningful edit restarts learning.
Why does the order matter?
Because starting with audience and budget means starting with the most expensive and least likely cause. Measurement and creative explain most cases and cost less to check.
How long does it take?
About an hour per account once it is a habit. The first time takes longer because you are defining your thresholds.
Should I log what I changed?
Always, with the reason attached. Without a log, three weeks later nobody knows why an ad set was paused, and the decision gets made again from scratch.