Edited an ad set and tanked it? You reset learning.
Learning phase needs ~50 conversions in 7 days. Touch the wrong field and the counter goes back to zero — CPMs spike, delivery stutters.
What actually resets it:
— Budget change over ~20% up or down
— New creative, audience edit, optimization event swap, bid/cost cap change
— Pausing 7+ days
What does NOT reset it:
— Budget moves under 20%
— Ad name, ad set name, schedule end-date tweaks
So scale spend in 15-20% steps, wait 24h, repeat. Slow ramp keeps you OUT of re-learning. Try the stair-step on one ad set tonight.
Learning phase needs ~50 conversions in 7 days. Touch the wrong field and the counter goes back to zero — CPMs spike, delivery stutters.
What actually resets it:
— Budget change over ~20% up or down
— New creative, audience edit, optimization event swap, bid/cost cap change
— Pausing 7+ days
What does NOT reset it:
— Budget moves under 20%
— Ad name, ad set name, schedule end-date tweaks
So scale spend in 15-20% steps, wait 24h, repeat. Slow ramp keeps you OUT of re-learning. Try the stair-step on one ad set tonight.
Your dupes keep dying. Here's why.
Most buyers dupe winners back into the SAME ABO (ad set budget optimization — you set spend per ad set). New ID, fresh learning, fresh competition with your own old ad set. Cannibalism.
Flip it:
— Dupe the winning ad set INTO a clean CBO (campaign budget optimization — Meta splits the budget for you) with 3-4 copies
— Set the CBO at 2x your current daily spend, not 10x
— Kill the original ABO 48h later once the CBO proves out
You stop bidding against yourself. The auction reads it as one strong signal, not two weak ones.
Try it on ONE winner this week. Not your whole account.
Most buyers dupe winners back into the SAME ABO (ad set budget optimization — you set spend per ad set). New ID, fresh learning, fresh competition with your own old ad set. Cannibalism.
Flip it:
— Dupe the winning ad set INTO a clean CBO (campaign budget optimization — Meta splits the budget for you) with 3-4 copies
— Set the CBO at 2x your current daily spend, not 10x
— Kill the original ABO 48h later once the CBO proves out
You stop bidding against yourself. The auction reads it as one strong signal, not two weak ones.
Try it on ONE winner this week. Not your whole account.
New ad account? Don't launch at $500/day.
Cold accounts with no trust history get flagged or throttled when you blast big spend day one. Worst case: cooked account (banned/restricted with no warning).
Warm it like a real business:
— Days 1-3: $30-50/day, simple compliant offer, clean creative
— Days 4-7: climb 20-30%/day if delivery is smooth and no flags
— Only push aggressive once you've got a payment + spend history
A week of patience beats a 281-day appeal.
Also: never run your best creative on a brand-new account. Save it for after warmup. Warm ONE new account properly before you trust it with real budget.
Cold accounts with no trust history get flagged or throttled when you blast big spend day one. Worst case: cooked account (banned/restricted with no warning).
Warm it like a real business:
— Days 1-3: $30-50/day, simple compliant offer, clean creative
— Days 4-7: climb 20-30%/day if delivery is smooth and no flags
— Only push aggressive once you've got a payment + spend history
A week of patience beats a 281-day appeal.
Also: never run your best creative on a brand-new account. Save it for after warmup. Warm ONE new account properly before you trust it with real budget.
Stuck in 'Learning Limited' forever? You're editing it back in every time.
Learning Limited means the ad set can't hit 50 conversions in 7 days. Every significant edit restarts the 7-day clock from zero.
Break the loop:
— Stop touching it. No budget, no creative, no audience tweaks for a full 7 days
— If it still can't hit 50, consolidate: merge it into a sibling so events pool
— Broaden the audience or move the event up the funnel (purchase to add-to-cart)
— Only THEN judge cost
Risk: pooling events into the wrong ad set can dilute a clean winner.
Pick ONE Learning Limited set and go hands-off 7 days. Hardest part is doing nothing.
—
Для любителей cpm benchmarks — @pixelprofit_fb
Learning Limited means the ad set can't hit 50 conversions in 7 days. Every significant edit restarts the 7-day clock from zero.
Break the loop:
— Stop touching it. No budget, no creative, no audience tweaks for a full 7 days
— If it still can't hit 50, consolidate: merge it into a sibling so events pool
— Broaden the audience or move the event up the funnel (purchase to add-to-cart)
— Only THEN judge cost
Risk: pooling events into the wrong ad set can dilute a clean winner.
Pick ONE Learning Limited set and go hands-off 7 days. Hardest part is doing nothing.
—
Для любителей cpm benchmarks — @pixelprofit_fb
Throwing away likes every time you dupe? Stop.
When you dupe an ad, Meta spins a fresh post ID — zero comments, zero shares, zero social proof. You rebuild trust from scratch every single time.
Use existing post ID instead:
— Publish your creative as a page post, grab the post ID (under Page > Meta Business Suite > posts)
— In every new ad set, choose "Use Existing Post" and paste that ID
— All engagement stacks onto ONE asset across all your scaling ad sets
300 comments of social proof beats a naked ad in cold traffic. Costs you nothing.
Warning: one report on that post nukes every ad using it. Keep a backup ID warm.
When you dupe an ad, Meta spins a fresh post ID — zero comments, zero shares, zero social proof. You rebuild trust from scratch every single time.
Use existing post ID instead:
— Publish your creative as a page post, grab the post ID (under Page > Meta Business Suite > posts)
— In every new ad set, choose "Use Existing Post" and paste that ID
— All engagement stacks onto ONE asset across all your scaling ad sets
300 comments of social proof beats a naked ad in cold traffic. Costs you nothing.
Warning: one report on that post nukes every ad using it. Keep a backup ID warm.
Cost cap killing your delivery? You set it too tight.
Fresh cost cap campaigns (you tell Meta the max avg cost per result) often get zero spend on day one. Buyers panic and raise it 3x — now you're overpaying.
Seed it loose, then choke it down:
— Launch the cap ~20-30% ABOVE your true target CPA
— Let it spend and exit learning (~50 conv)
— Drop the cap in 5-10% steps every 2-3 days toward your real target
Meta's model needs room to find conversions before you squeeze. Squeeze first = no data, no delivery.
Do this on ONE cost-cap campaign. Watch it actually spend.
Fresh cost cap campaigns (you tell Meta the max avg cost per result) often get zero spend on day one. Buyers panic and raise it 3x — now you're overpaying.
Seed it loose, then choke it down:
— Launch the cap ~20-30% ABOVE your true target CPA
— Let it spend and exit learning (~50 conv)
— Drop the cap in 5-10% steps every 2-3 days toward your real target
Meta's model needs room to find conversions before you squeeze. Squeeze first = no data, no delivery.
Do this on ONE cost-cap campaign. Watch it actually spend.
Hit a scaling cliff at $1.5k/day? You went vertical too soon.
Scaling cliff = the point where pumping more budget into one ad set just inflates CPA. Vertical scaling (more $ same ad set) hits it fast.
Go horizontal first:
— Duplicate the winner across 3-4 NEW audiences (lookalikes, interest stacks, broad)
— Each at the SAME budget the original liked
— Let Meta find fresh pockets of the auction instead of saturating one
Then, and only then, ramp the survivors vertically 15%/day.
Width buys you headroom. Height alone burns frequency. Test ONE horizontal expansion before you touch budgets again.
Scaling cliff = the point where pumping more budget into one ad set just inflates CPA. Vertical scaling (more $ same ad set) hits it fast.
Go horizontal first:
— Duplicate the winner across 3-4 NEW audiences (lookalikes, interest stacks, broad)
— Each at the SAME budget the original liked
— Let Meta find fresh pockets of the auction instead of saturating one
Then, and only then, ramp the survivors vertically 15%/day.
Width buys you headroom. Height alone burns frequency. Test ONE horizontal expansion before you touch budgets again.
Pairs well with this channel
@FeedHeretic — Debunking LinkedIn 'guru' advice with what the feed actually rewards — why broetry,… Quietly one of the better feeds in the space.
@FeedHeretic — Debunking LinkedIn 'guru' advice with what the feed actually rewards — why broetry,… Quietly one of the better feeds in the space.
CTR sliding but frequency is fine? It's not fatigue.
Everyone blames "creative fatigue" (audience saw it too much). But if frequency is under ~2.5 and CTR still drops, your problem is the FIRST FRAME, not overexposure.
Diagnose before you reshoot:
— Check 3-sec video view rate — falling means the hook died, not the ad
— Swap ONLY the first 1.5 seconds, keep the body
— Test a pattern-interrupt opener: motion, face, text-on-screen at frame one
Reshooting the whole ad when you just need a new hook wastes a week. Re-cut the intro, relaunch as a fresh post ID, measure thumb-stop. Try it on your tired top ad.
Everyone blames "creative fatigue" (audience saw it too much). But if frequency is under ~2.5 and CTR still drops, your problem is the FIRST FRAME, not overexposure.
Diagnose before you reshoot:
— Check 3-sec video view rate — falling means the hook died, not the ad
— Swap ONLY the first 1.5 seconds, keep the body
— Test a pattern-interrupt opener: motion, face, text-on-screen at frame one
Reshooting the whole ad when you just need a new hook wastes a week. Re-cut the intro, relaunch as a fresh post ID, measure thumb-stop. Try it on your tired top ad.
Test in ABO, scale in CBO. Here's the clean handoff.
ABO (you control spend per ad set) = great for fair testing. CBO (Meta allocates) = great for scaling winners. The mistake is scaling in the same structure you tested in.
The handoff:
— Test 4-5 ad sets in ABO at equal budgets, find your 1-2 winners by CPA
— Build a fresh CBO, drop ONLY the proven winners in
— Start the CBO at the COMBINED budget the winners earned, not higher
— Let it run 48-72h before any vertical push
Clean separation = clean data. No mixing dead tests into your scaler.
Try the handoff on your next winning test.
ABO (you control spend per ad set) = great for fair testing. CBO (Meta allocates) = great for scaling winners. The mistake is scaling in the same structure you tested in.
The handoff:
— Test 4-5 ad sets in ABO at equal budgets, find your 1-2 winners by CPA
— Build a fresh CBO, drop ONLY the proven winners in
— Start the CBO at the COMBINED budget the winners earned, not higher
— Let it run 48-72h before any vertical push
Clean separation = clean data. No mixing dead tests into your scaler.
Try the handoff on your next winning test.
Account spends great by day, garbage by evening? Dayparting.
CPAs often drift up after 6pm as cheaper inventory dries out and you compete with everyone's evening push.
Before you touch ad spend, look at the time breakdown:
— Ads Manager > Breakdown > By Time > Hour
— Find the hours where CPA is 40%+ above your average
— On CBO you can't pause hours cleanly, so use an ABO + ad schedule to mute the worst block
Don't kill a whole campaign over 3 bad hours.
Pull your hourly breakdown tonight. You'll probably find one ugly window eating margin.
CPAs often drift up after 6pm as cheaper inventory dries out and you compete with everyone's evening push.
Before you touch ad spend, look at the time breakdown:
— Ads Manager > Breakdown > By Time > Hour
— Find the hours where CPA is 40%+ above your average
— On CBO you can't pause hours cleanly, so use an ABO + ad schedule to mute the worst block
Don't kill a whole campaign over 3 bad hours.
Pull your hourly breakdown tonight. You'll probably find one ugly window eating margin.
Ramping spend? The 20% rule that keeps learning intact.
Big budget jumps (50%+ overnight) throw the ad set back into learning — delivery resets, CPMs jump, results wobble for days.
The disciplined ramp:
— Increase budget no more than ~20% per move
— Wait a full 24h between moves so the algo re-stabilizes
— If CPA holds, repeat. If it spikes, hold flat one extra day before deciding
Yes it's slower than doubling overnight. It's also why your account doesn't randomly faceplant on Tuesday.
Set a calendar reminder and ramp ONE campaign by the book this week. Compare it to your usual yank-the-budget move.
Big budget jumps (50%+ overnight) throw the ad set back into learning — delivery resets, CPMs jump, results wobble for days.
The disciplined ramp:
— Increase budget no more than ~20% per move
— Wait a full 24h between moves so the algo re-stabilizes
— If CPA holds, repeat. If it spikes, hold flat one extra day before deciding
Yes it's slower than doubling overnight. It's also why your account doesn't randomly faceplant on Tuesday.
Set a calendar reminder and ramp ONE campaign by the book this week. Compare it to your usual yank-the-budget move.
Running 1%, 2%, 5% lookalikes in separate ad sets? Stack them.
Separate lookalike (LAL — audience that mimics your buyers) ad sets fragment your budget and overlap badly — you bid against yourself in the 1-3% zone.
Consolidate:
— Build ONE 0-5% stacked lookalike audience in a single ad set
— Or go 1-10% if your seed list is small (under 1,000 buyers)
— Let CBO distribute, don't slice it into thin parallel sets
Fewer, fatter ad sets exit learning faster and stop overlapping.
Refresh the seed list every 2 weeks so the model tracks your real buyers. Try one stacked LAL against your sliced setup.
Separate lookalike (LAL — audience that mimics your buyers) ad sets fragment your budget and overlap badly — you bid against yourself in the 1-3% zone.
Consolidate:
— Build ONE 0-5% stacked lookalike audience in a single ad set
— Or go 1-10% if your seed list is small (under 1,000 buyers)
— Let CBO distribute, don't slice it into thin parallel sets
Fewer, fatter ad sets exit learning faster and stop overlapping.
Refresh the seed list every 2 weeks so the model tracks your real buyers. Try one stacked LAL against your sliced setup.
CBO dumping all budget into ONE ad set? Use min/max caps.
CBO loves to feed its favorite child and starve the rest — fine until that ad set fatigues and the whole campaign tips over.
Force diversification:
— Open ad set > advanced > set a minimum daily spend on your secondary sets
— Or cap the max on the greedy one so it can't hog everything
— Keep 2-3 ad sets genuinely funded so you have backups when the leader dies
This trades a little efficiency for a lot of stability.
Warning: over-restrict and you fight the algo for nothing. Use light touches. Try a minimum on ONE backup ad set in a fragile campaign.
CBO loves to feed its favorite child and starve the rest — fine until that ad set fatigues and the whole campaign tips over.
Force diversification:
— Open ad set > advanced > set a minimum daily spend on your secondary sets
— Or cap the max on the greedy one so it can't hog everything
— Keep 2-3 ad sets genuinely funded so you have backups when the leader dies
This trades a little efficiency for a lot of stability.
Warning: over-restrict and you fight the algo for nothing. Use light touches. Try a minimum on ONE backup ad set in a fragile campaign.
Killing ads after 1 day? You're murdering winners early.
Judging on time ("24h and no sale, dead") ignores that learning is volume-based, not clock-based.
Kill on spend, not the calendar:
— Give each new ad ~1.5-2x your target CPA in spend before judging
— $40 CPA target? Let it spend $60-80 before the verdict
— No add-to-carts or landing views by then = genuinely cut it
Low-volume accounts especially need this — a single day means almost nothing statistically.
Flip side: don't let a true dud run forever "to be fair." Set the spend threshold up front. Apply it to your next batch of new ads.
Judging on time ("24h and no sale, dead") ignores that learning is volume-based, not clock-based.
Kill on spend, not the calendar:
— Give each new ad ~1.5-2x your target CPA in spend before judging
— $40 CPA target? Let it spend $60-80 before the verdict
— No add-to-carts or landing views by then = genuinely cut it
Low-volume accounts especially need this — a single day means almost nothing statistically.
Flip side: don't let a true dud run forever "to be fair." Set the spend threshold up front. Apply it to your next batch of new ads.
Prospecting CPA creeping up week over week? Check frequency.
On cold campaigns, frequency over ~2.5-3 in 7 days means you're re-hitting the same people and paying more for worse results.
Reset the pool without nuking the campaign:
— Pull frequency: Breakdown isn't needed, it's a column you can add
— If it's climbing, inject a NEW audience or new creative angle to widen reach
— Or move spend to broad to escape the saturated narrow pocket
Don't just raise budget on a saturated audience — that ACCELERATES frequency and burns you faster.
Add the frequency column to your prospecting view today. Watch the trend, not just the snapshot.
On cold campaigns, frequency over ~2.5-3 in 7 days means you're re-hitting the same people and paying more for worse results.
Reset the pool without nuking the campaign:
— Pull frequency: Breakdown isn't needed, it's a column you can add
— If it's climbing, inject a NEW audience or new creative angle to widen reach
— Or move spend to broad to escape the saturated narrow pocket
Don't just raise budget on a saturated audience — that ACCELERATES frequency and burns you faster.
Add the frequency column to your prospecting view today. Watch the trend, not just the snapshot.