Way back in the second issue of this newsletter (#002), back in April, taught the Daily Brief Agent. It scraped several YouTube channels, Twitter profiles, and a couple newsletters saving me 3-4 hours a week of staying up to date on content in my market.
Useful. But it was training wheels stuff.
We are going to build off of that knowledge today and create what I've checked first every morning for almost 20 years: ad spend.
Log in, squint at yesterday, decide if I'm bleeding. Every media buyer alive runs that same ritual, and most of us have run it so long we've never stopped to ask if the number we're squinting at is even done yet.
It isn't. And that's the problem this whole build exists to fix.
So welcome to Part 1 of the AI Paid Ads Department.
What you're about to build is not the newbies version of an ad dashboard. If you're thinking about skipping this because you've heard this all before, lol, brother... stay tuned.
I may be an AI guy now, but I was the silverback gorilla in paid ads.
This month we're building one complete department across four Thursdays, and we start with the eyes: the rebuild of #002, except this one reads your ad accounts before you're out of bed, and it knows something about your dashboard that you probably don't.
The Wet Paint Rule
Your dashboard shows you one number for yesterday. It does not tell you that some of yesterday is FINISHED and some of yesterday is still being painted.
Painted??? What the...
Spend, clicks, and impressions? Those are dry. Google refreshes them on an hourly cycle, and Meta refreshes its reporting every 15 minutes. What you see this morning is for sure what happened.
Conversions are a different animal.
Google files a conversion under the day of the CLICK, not the day of the sale. Somebody clicks your ad Tuesday, thinks about it, and buys Friday... that sale gets written into TUESDAY's row.
You following me? Conversions take time to show up, aka "wet paint."
Google's own documentation says conversions can be reported up to 90 days after the click (and after significant account changes, it can take up to 7 days for conversion data to fully show up at all).
Meta's version: insights refresh every 15 minutes but can keep restating for up to 28 days before the number locks. Which means last Tuesday's conversion number isn't a fact. It's a rough draft that keeps improving for weeks.
Now run the ritual again. A marketer looks at yesterday, sees 3 conversions where last week had 7, and feels that cold drop in the stomach. Know what that number was actually doing?? Still filling in. They pause the campaign anyway, judging paint that was still WET, and three days later the row fills in to 9.
The campaign they just killed was their best one.
Look you don't learn this stuff watching a coupla YouTube vids. I've been doing paid ads for 20yrs, and spoke all over the world about it. But AI is my new playground, I'm just bringing my past experience into this new world.
I call the fix The Wet Paint Rule.
Never render a verdict on a number that's still filling in. Alert on spend today, judge conversions on a lag. Your dashboard presents wet numbers and dry numbers in the same font, and that single design choice has burned more ad budget than bad creative ever did.
The Four Rules of Not Fooling Yourself
The Wet Paint Rule has three siblings, and together they're the entire discipline of reading an ad account like a professional instead of a nervous gambler:
1. Compare like windows. Rolling 7 days against the prior 7 days. Never Tuesday against Monday, because Tuesday always loses to Monday in some industries and always wins in others. Day-of-week rhythm is real, and single-day comparisons can't tell rhythm from trouble.
2. Let conversions dry before judging. Exclude the trailing 3 days from any conversion verdict on Google, and the trailing day on Meta. The spend data is trustworthy same-day. The conversion data is not. Different confidence, different treatment.
3. Gate alerts on volume, not percentage. A campaign that drops from 2 conversions to 1 is "down 50%," and it means NOTHING. Percentages lie at low volume. Set a floor (I use 20 conversions in the window) before a change is allowed to become an alert.
4. Mute anything still learning. Google says a bid strategy can take up to 3 weeks or 1 to 2 conversion cycles to calibrate after a change. Meta has its own learning phase after edits. Performance inside that window is EXPECTED to be unstable. An agent that panics during learning is an agent that trains you to ignore it.
Those four rules are simple to say and miserable to practice manually, every morning, across two platforms, before coffee. Which is exactly the job description of an agent.
What Good Even Looks Like
One more thing the brief needs before it can grade anything: a grading scale.
WordStream published their 2026 benchmarks in May, built from 13,474 US search campaigns: median click-through 6.64%, median cost per click $5.42, median conversion rate 8.18%, median cost per lead $66.69.
Those medians are your floor, not your goal.
My June numbers, pulled by the Profit Dashboard Agent from #015, had my cost per lead at $9.87 against that $66.69 median. That gap isn't magic, it's just attention to detail... the kind this agent hands you every morning for free.
So the brief grades every campaign twice: against YOUR targets first, industry medians second. A $40 CPL is a party in one business and a house fire in another. The agent has to know which business it works for.
To pull this off you need a Google ads MCC account, developer token, Google ads MCP, Meta ads account, the Meta ads MCP, and research GAQL (Google ads query language).
Continue with Build Notes+
This is the free preview. The complete step-by-step guide and supporting files are available to Build Notes+ members.
