← All articles
Ad Intelligence5 min read

How to analyze Facebook ads performance weekly

How to analyze Facebook ads performance in a fixed weekly order: account, campaigns, ad sets, ads, then three written decisions. My Monday routine.

Monday, 9:40. Ads Manager is open on the last 7 days, your column preset holds 23 metrics, and you have been scrolling sideways for twenty minutes. CTR is down on four ads and up on three. ROAS on the main campaign slid from 2.6 to 2.3. Someone on the client side wants to know what we learned this week. You have plenty of numbers and no sentence.

I have run that Monday more times than I can count. The fix was never a better dashboard. It was an order: read the account from the top down, weight everything by spend, and keep the tab open until you have written three decisions.

Short answer: To analyze Facebook ads performance, work top down in a fixed weekly order. First check account-wide spend, CPM and conversion rate to rule out tracking or market moves. Then see where campaign and ad set budget went, review ads weighted by spend, and finish with three written decisions: what to scale, cut and test next.

The takeaways

  • Rule out the account before you blame an ad. If CPM and conversion rate moved on every campaign in the same week, the cause is the auction or your tracking, and no single creative explains it.
  • Weight the ad review by spend. Sort by amount spent and stop once you have covered about 80 percent of the week's budget; a 6 € ad has nothing to teach you yet.
  • The output is three decisions. One thing to scale, one to cut, one to test next, each written down with the number that justified it.

Why start with the whole account instead of the ads?

Most week-over-week movement in a Meta account is account-wide, and an account-wide move cannot be explained by any one ad. Compare this week with last week on three numbers only: total spend, CPM and conversion rate (purchases or leads divided by link clicks). If all three moved together across campaigns, look outside the creatives first.

A CPM jump on every campaign in late November is the auction getting crowded before Black Friday. A conversion rate that drops everywhere at once while CTR holds points past the ad: a checkout change, a slower landing page, or a pixel that stopped deduplicating against the Conversions API. Check that the week's purchases in Ads Manager roughly match your shop backend. If the two drift apart, fix measurement before you analyze anything else, because every number further down inherits the error.

If the week fell off a cliff, switch to the diagnosis order for a sudden drop. This routine is built for an ordinary week.

Where did the budget go at campaign and ad set level?

The campaign and ad set pass answers one question: did the money go where you intended, and did it buy results there? Sort campaigns by spend and put each one's share of spend next to its share of conversions. A campaign taking 40 percent of spend for 20 percent of purchases is your first conversation of the week.

At ad set level, check three things. Is any ad set still in learning, or marked "Learning limited" in the delivery column? Its numbers are unstable and should not drive a budget call yet. Did frequency creep up on retargeting ad sets? Small audiences saturate first. And did an automated rule or a colleague change budgets mid-week? The activity log shows it, and a mid-week change means this week's numbers blend two setups.

Write down what you find and leave the budgets alone until you have seen the ads.

How do you review individual ads without drowning in them?

Sort ads by amount spent, descending, and work down until you have covered roughly 80 percent of the week's spend. Everything below that line has not spent enough to judge. A 6 € ad with a 9× ROAS is one lucky purchase.

For the ads above the line, read the metrics in funnel order: hook rate for video, then CTR, then cost per result, then ROAS. I wrote up why that reading order matters separately. A weak hook with a healthy CPA means the people who stop are qualified and you need more of them stopping. A strong CTR with a bad CPA points at the landing page or the offer.

The trap at this level is small counts. An ad with 3 purchases this week and 7 last week has not halved in quality; at those counts the swing is ordinary noise. Compare each ad against its own last 14 days and against what its format usually does, then ask whether the change is bigger than the noise.

What should a weekly Facebook ads analysis produce?

A weekly Facebook ads analysis should produce three written decisions: one thing to scale, one thing to cut, and one test to run next, each with the number behind it. A review that ends in a screenshot of metrics was reporting. The decisions are the deliverable, and they are what you check against next Monday.

Here is the format I use, with illustrative numbers: "Scale: UGC hook v3, cost per purchase 31 € over 2,400 € spend, raise budget 20 percent. Cut: static carousel B, frequency 4.1 and CTR down three weeks running. Test: the price-anchor angle from v3 as a static."

Next week you open that note first. Did the scaled ad hold its cost after the raise? Did the cut free money for something better? After a month the note is the only honest record of whether your calls were any good.

Some weeks the right decision is to change nothing. Write that down too, with the reason.

Which parts of the routine can software take over?

Software can take the arithmetic of this routine off your hands, and the three decisions stay with you. I built Adscalr around that split. Its composite score combines six metrics (hook rate, CTR, CPI, ROAS, share rate, revenue per install) and pulls a new ad's early score toward what its format usually does, so the lucky 6 € ad does not top the list. Fatigue detection reads a multi-day CTR slope and checks it against the whole market, which covers part of the account-level step. A role-based morning brief does the daily first pass.

None of that explains why an ad works, and none of it writes your Monday note. If you want to see how the scoring handles small samples, the ad intelligence page walks through it.

This is the thinking behind Adscalr.

See the product →