If you want to see a rainbow you have to learn to see the rain.
Joined August 2011
- Tweets742
- Following1.1K
- Followers235
- Likes2K
Milan Cidilo retweeted
Evaluating Performance Max on aggregate ROAS and CPC is the fastest route to bad decisions.
A real example from our data:
Looking at the cost to conversion value ratio for YouTube and Discover in a PMax campaign, they looked almost identical. Similar profitability, similar ROAS.
When you break them down in our PMax script by clicks and conversion rate, reality looks different:
YouTube:
• Very low clicks
• High CPC
• High conversion rate
Discover:
• High click volume
• Cheap traffic
• Low conversion rate
Looking at the campaign average hides how each channel behaves.
This is also why the new Channel Prioritization feature is limited. You can cap Search Partners or Display, but YouTube, Discover, Gmail, Maps, Search, and Shopping stay in one indivisible group.
The algorithm still decides where to spend, and you cannot set a cap on individual channels inside the interface.
The new feature is fine for restricting Search Partners, but deep analysis still requires tracking channel metrics over time.
Read how to get the most out of the script and where it helped us: get-pmaxscript.6clickz.com/
Milan Cidilo retweeted
Claude turned 1,256 competitor ads into my next creative brief.
349 survived 30+ days. The other 907 were switched off.
8 Meta Ad Library searches became a complete map of what my market is testing.
→ No more $5,000/mo creative research retainers
→ No more brainstorming from a blank whiteboard
→ No more guessing which hook deserves budget
Just 8 searches in → 1 evidence-backed creative brief out.
Here's how it works:
→ Scan (8 searches → 863 clean ad records)
→ Survival check (1,256 versions → the 349 still running after 30 days)
→ Compare (245 advertisers → the 14 worth breaking down)
→ Transcribe (478 videos → 387 unique scripts)
→ Break down (every survivor → hook, visual, offer and CTA)
→ Brief (the patterns → 1 original creative concept)
Every idea keeps the original ad, advertiser and source attached. Inspect the evidence instead of trusting an AI guess.
Included: the dashboard, all 9 Claude skills, the full pipeline and a worked run.
Want the complete AI Creative Research Engine?
Like + reply "ENGINE" + repost, and I'll DM it to you.
When e-commerce sales stall, media buyers reflexively rebuild ad creatives.
In most stores, the drop-off happens between the product page and checkout.
The ad earned the click. On-site UX friction determines whether they convert.
Milan Cidilo retweeted
i have insider info:
claude duet is f*cking goated for running meta ads and printing millions, for literally any business.
it’s still behind a waitlist, but from what we’re hearing it briefs you every morning, watches your ads and leads all day and does the work while you’re away.
no media buyer, no VA, no agency.
so we built the whole setup before it drops, and it already works in claude today.
free for the next 24 hours:
• the 7am account brief: spend, cost per lead vs target, and anything that broke overnight
• the fatigue alarm that flags a winner before it dies (our real rotation rules)
• the kill check: every ad that’s spent 2x your target with no result, waiting for your OK
• the speed to lead watchdog: pings you when a lead hasn’t been called in 5 minutes
• the swipe watch: competitor ads that just passed 90 days running
• call transcripts and reviews turned into 10 different ad concepts
• 5 replacement ads drafted the morning a winner fatigues
• the friday report, written and waiting for your approval
• the approval rulebook: what it does alone and what it never touches
• the whole thing on one whiteboard you can read in a minute
backed by $12M+ generated and over a million a month in managed spend.
like + comment “DUET” and I’ll send it over
(must be following so I can DM you)
repost for priority
Switching a winning static into a UGC reel is not a new test concept.
It's a format iteration on the exact same message.
Valuable for managing placement CPMs. Completely useless for breaking through audience saturation ceilings.
Concept = Persona × Angle × Format
• Persona dictates the auction segment (who enters).
• Angle dictates persuasion (why they convert).
• Format dictates feed placement (CPM efficiency).
If you only change format and hook, you never unlock new reach.
Changing custom label logic in PMax drove +133% conversions and +95% revenue without spending an extra euro.
Most DTC brands have hundreds of SKUs starved of budget by the algorithm. Fixing the feed structure unlocked that dormant inventory.
To a designer, a concept is a layout.
To a copywriter, it is a headline.
To a media buyer, it is an MP4.
When your team defines the unit differently, you ship 30 ads that Andromeda clusters into one.
Align on the atomic unit before you open the edit bay.
Sorting competitor ads by impressions in Meta Ad Library is a trap.
Ad Library impressions reflect where someone allocated budget, never whether the creative converted profitably.
Copying high-impression ads usually just means copying another brand's unvalidated test spend.
Testing 20 hook tweaks of the same benefit is cosmetic, not strategic.
A real creative concept requires three distinct variables:
Persona × Angle × Format
Change only the format, and you're just paying production tax on the exact same auction.
Meta decides delivery sequence.
Your job is to supply concepts far enough apart that the model can build completely distinct delivery paths.
Volume creates activity. Message distance creates incremental reach.
The Stagnation Paradox:
Spending more on creative production while unique reach flatlines.
It happens because Meta's sequence learning model doesn't just rank single ads. It maps the full user journey.
If every ad targets the same angle, the algo simply has nowhere new to go.
Doubling your creative output often leads directly to flatlining unique reach.
When you feed Meta 15 minor hook variations of the same idea, Andromeda clusters them together and GEM applies a similarity discount.
You paid for 15 ads. The ranking model treats them as one.
Most creative strategists never graduate past script volume.
The real shift is measuring hit rate by bucket. At 6clickz we split 144 concepts into big swings, new concepts, and iterations.
A blended hit rate hides which bucket actually carries the account.
Milan Cidilo retweeted
An account doing 53% of last year's revenue in Q1 is not automatically a failing account.
On Dedoles that quarter was the plan working. By Q4 the same account was at 102%.
ROAS followed the same shape. Q1 at 73% of the previous year, Q4 at 121%.
The context matters. One of the largest Slovak e-commerce projects, in restructuring, moving off TV advertising onto performance marketing with hard KPIs. CPC and CPM were up year on year across every platform. The brief was to match last year's revenue and spend less doing it.
Q1 went on taking inefficient spend out. Revenue dropped, because that is what happens when you stop paying for the campaigns that were carrying the number.
Q2 and Q3 rebuilt volume at a lower click price.
October turned out to be the best month of the year. Revenue matched the previous year, ROAS came in 37% better.
Anyone reviewing this account on a monthly cadence would have stopped it in March.
Rule: a turnaround is shaped like a dip. Agree on the quarter you will be judged in before touching the first campaign, because the dip is where the decisions that produce Q4 get made.
Have you had to defend a planned revenue drop to a board? How did that land?
Read more: dedoles-en.6clickz.com/
Milan Cidilo retweeted
The 80/20 of the PMax channel split:
1. Cost and conversion value: who spends the money, who brings the value back
2. Clicks and conversions: why your blended numbers look wrong
3. Read it by day, not by period, or you find out a week late
Discover buys clicks. Shopping buys conversions. Averaged together they describe neither.
Milan Cidilo retweeted
A bad PMax conversion rate next to your Search campaigns means nothing on its own. You are not comparing two campaigns. You are comparing one campaign against a mix of channels that happens to sit under one name.
Same for CPC that looks suspiciously cheap. Same for a ROAS from another planet.
The mechanism is boring once you see it:
> Discover pulls a pile of clicks and converts poorly
> Shopping takes fewer clicks and converts well
> Blend them and you get a conversion rate that describes neither
So before you judge a PMax campaign against anything else in the account, split it. Two charts, two questions.
Cost and conversion value on the first one. Who spends the money, who brings the value back. That is your baseline.
Clicks and conversions on the second. This is where you find out why the blended numbers look the way they do.
The stacked version of this is the one thing I still miss in the Google Ads reports, so it stayed in our script.
Judge PMax against its own channel mix first. Against the rest of the account second.
Free PMax Script, and on the same page a PDF with the six cases where it saved or made real money on client accounts: get-pmaxscript.6clickz.com/
Milan Cidilo retweeted
Claude is INSANE for ad creative strategy.
I built 5 Claude skills that handle the entire creative pipeline.
Not prompts you paste and forget. Skills you install once and they run automatically.
→ Full audience breakdown with real pain points and buying triggers
→ Ranked ad angles tied to different funnel stages
→ Production-ready copy in the buyer's voice
→ Visual direction specific enough to hand straight to design
→ Batch variations across hooks, tones, and formats
5 skills. One terminal. Finished ad creatives in minutes.
Prompts are one-time. Skills compound.
They carry context from every run. The strategy gets sharper every time you use them.
If you want the 5 skills:
Connect with me
Comment "MAP"
I'll send it to your DMs.
Repost for priority access.
Milan Cidilo retweeted
a human media buyer works 8 hours a day.
my AI media buyer works 24/7.
while i sleep, my agentic Claude Opus 4.8 loop is checking our creatives every 15 minutes through the Meta’s MCP.
it automatically scales winners / pauses losers so I NEVER lose money from poor management, ad fatigue or forgetting to increase bids
so i documented the ENTIRE playbook on how you can do the same…
here’s what’s included inside the guide:
1. setup - Meta's MCP connected to Claude:
(system user token + ads_management perms, JSON config block, verify connection. Claude OAuth or Claude Code config for multi-account)
2. the master operating prompt:
(monitoring cadence every 15 min at peak, decision authority, reporting format with a live account health score)
3. the decision rules:
(creative fatigue at CPM up 20% w/w, CTR down 15%, thumbstop under 25%, frequency over 3.0)
3. the 15-minute ROAS shift detection prompt:
(trailing 3-hour vs 7-day ROAS, real-time reallocation capped at 15%)
4. hook performance segmentation:
(sorted by naming convention - AUTH, CONT, PAIN, PROOF - to see which angle carries the spend)
5. creative rotation workflow:
(min 10 approved creatives tagged BACKUP, rotated like-for-like, plus the rotation prompt)
6. what it doesn't replace:
(strategy, offer, funnel logic still come from you. a force multiplier on a good strategy, not a fix for a broken one)
all backed by $10M+ generated for service-based businesses through AI meta ads funnels
and for 24h you can access this entire system for FREE
like + comment "MEDIA" and I’ll send it over.
(must be following + RT for priority access)
