Beyond the Boost Insider
Smarter ads. Practical AI. Real growth.
Nine days to Black Friday, {{First Name | Friend}}
You locked your campaign budgets last week. You scheduled your emails. You set up your retargeting audience. You uploaded your creative. You can see it all running right now. If you haven’t and need support, please don’t be shy and reach out! No strings attached. I can help you out!
And every operator instinct you have is screaming at you to tweak something.
This week’s Boost is about the hardest part of the BFCM plan I gave you last Tuesday. The part that trips up more brands than any ad targeting problem ever does.
Discipline. 🤯
Here is what I have learned running Meta ads through 10+ Black Fridays: the campaigns that scale during BFCM are run by operators who can sit on their hands. Patience is the strategy.
THE BOOST: Discipline Is the Strategy
Why your hands have to stay off the dashboard
Every edit you make to a live ad set resets what Meta calls the learning phase. That reset takes 48 to 72 hours of spend to work through. If you edit your top-performing ad set on Nov 24, Meta is still “learning” through Black Friday. Can you imagine???
Your campaign stops being optimized at the exact moment it should be printing.
The metric that looks scary (but is normal):
Day-over-day CPM. From Nov 15 through Nov 25, CPMs rise across the entire Meta auction every single year. Every brand in your niche is bidding more aggressively. Your CPMs will go up 30 to 60 percent. That is the auction, not your account. Your targeting is fine.
The metric that actually tells you something needs attention:
Your 7-day rolling ROAS trend. If this drops below your baseline by 25 percent or more, AND your frequency is under 3 (meaning you are not saturated), you have a real signal worth paying attention to.
If frequency is over 5, you are saturating your audience. Note it in a doc for post-BFCM analysis. Do not touch.
The difference between those two numbers is the difference between a calm operator and a panicked one.
Three rules for the next 9 days:
If you absolutely must change a budget, change it by 20 percent or less, once per day. Anything above that triggers a learning reset.
Do not duplicate ad sets to “test something new.” You do not have 14 days to learn the new ad set. You just added noise to your account.
Do not add new audiences. Your warm audience is baked. Your cold audience is baked. Trust what you built two weeks ago.
What to watch, 10 minutes maximum, once a day:
Delivery (are your ads actually spending, or did one pause on a billing error)
Frequency (over 5 means you are saturating, note it, do not touch)
7-day rolling ROAS (the only performance number worth reading right now)
CAPI events in Events Manager (is purchase firing clean)
Close the dashboard. Open your post-click experience. Your product pages, your checkout, your abandoned cart flow. That is where every point of optimization lives for the next 9 days.
One thing about the data:
Meta does not report in real time. Conversions can take 24 to 72 hours to show up, and sometimes longer during BFCM when server loads peak. If your dashboard says ROAS 2.1 on Tuesday morning, that number is not final until Friday at the earliest.
The other complication is attribution. A customer can click your ad, close the tab, open your email three days later, and buy from the email link. Klaviyo gets credit. Meta looks like it underperformed. Your ad did the actual work of getting them to consider you.
What this means for the next 9 days: use your Shopify sales data, not Meta’s reported revenue, to decide if anything is actually wrong. If your Shopify revenue is tracking normal AND your 7-day rolling blended ROAS (total revenue divided by total ad spend across all channels) is holding, your ads are fine. Keep your hands off.
This is why the three-day lag is a trap. You check Monday’s number on Monday night, panic because it looks low, edit a campaign on Tuesday, reset the learning phase, and by Thursday you find out Monday was actually a good day once the delayed conversions reported.

