6 things I built in 6 days that used to cost me $40,000
Last month, we demoed what we built for Ginny Marvin.
If that name doesn't ring a bell: Ginny is Google's official Ads Product Liaison. Former Editor-in-Chief of Search Engine Land. Two decades in the industry. The human bridge between Google's product teams and every serious paid media professional on the planet.
Her reaction, verbatim:
"Oh my god. This is freaking awesome. How the hell did you do this?"
Prompts and a computer.
That is the complete technical answer.
Here's what she was looking at.
For fifteen years, my team at AdVenture assembled the cross-channel picture by hand.
Pull the Google data. Pull the Meta data. Pull the actual business numbers. Sit with our analytics lead Nechama and grind through the MMM outputs until the real story separated from the numbers. Figure out which anomaly was the truth and which one was the Facebook attribution window doing what attribution windows do — claiming credit they haven't earned, in windows they designed themselves.
Write down the judgment call. Try to remember it existed next month.
Every month. Without fail. For fifteen years.
Here's what that process produced, beyond the deliverable: a calibrated instinct. Which numbers to trust. Which swings to ignore. What a structural performance problem looks like versus a reporting artifact that disappears when you change the date range. That kind of judgment doesn't live in a dashboard. It lives in the people who built it — and that's what clients pay us for.
So we encoded it.
Google, Meta, actual revenue, actual margin, MMM, incrementality — one intelligence layer, built around the reasoning patterns our team developed across fifteen years of real accounts. The cross-channel picture now assembles itself. The month-end war room is over.
What we built is the institutional memory of an agency, made queryable.
That's what Ginny saw.
Then I kept going.
If fifteen years of accumulated judgment could be encoded in a matter of weeks, the obvious next question was — what else has been sitting in our heads that shouldn't be?
Turns out: a lot.
Over the next six days, I built tools I'd been meaning to commission from a developer for years. The quotes always came back somewhere between $8,000 and $28,000, with timelines measured in months, and a long back-and-forth about what I actually meant. So I never pulled the trigger.
A few of them:
A budget pacing alert that monitors spend trajectory across both platforms simultaneously, flags drift before it compounds, and tells you which specific campaigns to adjust. The check you currently run by instinct, a week after it would have mattered. ($3,000–$5,000 to build. I built it in a day.)
An audience overlap detector that maps every segment across every active campaign before anything goes live, identifies cannibalizing overlap, and surfaces specific resolution recommendations. The error you currently find during the debrief, after the spend has already exited the account. ($3,500–$5,000.)
A competitor intelligence tool. Drop in a domain. Four minutes later you have their creative approach, their messaging architecture, their offer structure, and the gaps they left open. The research your team does manually — in browser tabs nobody finishes. ($5,000–$8,000.)
A proposal generator built around our actual agency positioning. Scope goes in. A complete, formatted proposal in our voice comes out. Three hours of senior time, returned to your week, every time. ($4,000–$6,000.)
Rough midpoint on all of it: north of $40,000. Actual cost: the subscription I was already paying, and the hours it took to describe what I wanted with enough precision to get it built.
That last part matters. Describing a problem with precision — knowing exactly what it is, where it breaks, what a good output looks like — is a skill. It's not a technical skill. It's a practitioner skill. It's the thing paid media professionals have been developing their entire careers, inside real accounts, with real budgets, against real deadlines.
You already have it.
So let me ask you something directly.
What is the tool you've been meaning to build for two years?
The audit process that lives entirely inside one person's head. The onboarding workflow every new account manager rebuilds from scratch because it has always existed as tribal knowledge. The budget check that would have mattered last week. The brief that takes a senior person thirty minutes of careful conversation and should take no one any time at all.
The gap between having a clear idea and having a working tool used to require four things: a developer, a budget, a timeline, and a translation layer between what you meant and what got built.
All four are gone.
What remains is the problem-clarity that practitioners develop naturally — and the willingness to actually build the thing.
On Tuesday, March 24th, 10:00 AM to 5:00 PM Eastern, I'm running a full-day live Claude Code workshop on Zoom. You'll install it in the first hour. You'll build something real. You'll ship it to a live URL before we close.
No coding background. No developer. No prior technical experience.
The only prerequisite is a clear idea of what you want to build and the ability to describe it plainly.
You've already spent years developing the harder skill.
Register here. Tuesday is five days out.
Cheers,
Isaac
P.S. Ginny Marvin asked how. Now you know.
Responses