Going Live

When Nick accepted the Nashville proposal, the AI had a reasonable suggestion. Don't build a checkout. Just send him an invoice. He's your first customer. Your only customer (rubbing it in). You can figure out billing later.

I said no. If we're going to do this, we're doing it right. The first time someone says yes, they should be able to commit. Card in, campaign confirmed, through the same flow every future customer will use. Why wait? Let's do it now.

This was a principled decision that cost me about $25 in Stripe processing fees and whatever's left of my dignity.

The checkout that worked (and then didn't)

We built the whole thing. Nick's Nashville campaign page had the venue map, the strategy, the impression targets, and at the bottom, a checkout. Enter your card, confirm, you're in. We added a badge to the pricing: the $999 crossed out, $749 underneath with "Guinea Pig Rate" in Nick's own words.

That part went great. Nick entered his card. We collected the payment method. The plan was to start the subscription when the campaign launched, a date we hadn't even set yet.

Then we moved to setting up the subscription in Stripe. Somewhere in the process, the AI pushed back on "guinea pig rate" as a coupon name. Fair enough. If we're building this for future customers, "early adopter" makes more sense as a reusable discount. I didn't object.

The problem is the AI also decided that "early adopter" meant 15% off. Our deal with Nick was a flat $250 off. $749, not $849. And instead of scheduling the subscription for a future start date, it activated immediately.

Wrong amount. Wrong date. Charged to Nick's card under "Archway Ventures" because we hadn't updated the Stripe statement descriptor to Dropsignal yet. Our first-ever Stripe transaction was an $849 charge from a company name Nick had never heard of, for a campaign that hadn't started, at a price we didn't agree to.

We refunded it within three minutes. Customer email notifications were off in Stripe, so there's a decent chance Nick never noticed. But Stripe keeps processing fees on refunds. And my Stripe dashboard, the one I wanted to show my wife as proof that this thing was actually working, showed a negative balance. Customer number one, and we were in the red before we started.

The AI, to its credit, owned it. "That's on me." We deleted the subscription entirely, rebuilt it from scratch with the correct coupon ($250 flat, not 15%), the correct start date, and the correct statement descriptor. The second attempt was clean. And honestly, the screw-up made us go through every detail with the kind of care you should probably bring to billing from the start.

Designing the Nashville run

With the billing sorted, we could focus on the actual campaign. The plan for Nashville was built around a listener's day.

Morning commute: billboards along I-40, I-65, I-24. The big formats that catch you on the way to work. Mid-morning through lunch: office lobbies, the screens you walk past between meetings. Afternoon: gas stations, gyms, the errands and routines. Evening: bars, restaurants, the places you're choosing to be. Each venue type targeted to the time of day when someone is most likely to be in front of it and in the right headspace to notice a podcast recommendation.

We mapped 1,220 venues to this journey. Each one got a pin on the campaign map, color-coded by type. The strategy section described the rationale for each venue type. Not just where the screens are, but why that screen, at that time, for that listener.

Here's the part I'm most proud of: none of it was manual. From day one, we built automation for every step of the Vistar workflow. Campaign creation, venue targeting, creative upload, operator submission, delivery tracking. All through tools we built on top of the Vistar API. I never once logged into the DSP and manually created a plan, dragged creatives around, or hand-mapped venues to line items.

If the product works, we do this dozens of times. The first campaign was the hardest one to automate. Every edge case was new. But now when Nick's next episode drops, his creative refreshes automatically. Delivery data pulls every four hours and updates his dashboard. Creative rotation weights adjust so the newest episode gets the most airtime. The campaign runs itself.

Nick's response when I sent him the live dashboard link: "Cool, thanks for the heads up!" Not a lot of words. But he can click into that page anytime and see exactly what's happening in Nashville. Every venue, every day, every screen type. That felt like the product working.

Getting Brian back

Brian had gone silent again. I'd forwarded Nick's reply to show him his advice worked. I sent a follow-up asking if the email got stuck in his Superhuman spam, or if my ask for Spotify access was out of line, or maybe he decided to do the click farm thing instead, or maybe he was too busy writing his book and running Tacklebox and this was too big a distraction, or maybe he had other things going on that were none of my damn business. No reply to any of it.

So I went to his Calendly page. His paying members get a weekly Wednesday call. I'm not a paying member. The Wednesday slot had nothing available, but his main Calendly page showed all his event types. One of them was "Friday Coffees." Open. Not intended for me. I booked it anyway.

Brian's reply arrived the next morning: "Not sure why this let you book here — flipping calendly. I'm out of the office in the afternoon."

He blamed Calendly. He wasn't wrong that something had let me through. He was slightly wrong about what that something was.

I wrote back and told him exactly what I'd done. "I shouldn't have tried to sneak on your calendar via an event type that wasn't meant for me. I knew it and I did it anyway to get your attention. And I feel like a jerk. Sorry." Then sent the real email: we'd refined the approach since we last talked, used another podcaster's data to find markets based on where listeners aren't strong rather than where they already are, and if he wanted a head start, he could export his Spotify listener data as a CSV.

He replied an hour later. He'd had the flu and was teaching a Columbia course. Not an indicator of disinterest — interested customers can be the worst, he said, which is either very wise or a warning. April, when his new episodes start.

He still hasn't sent the CSV.

What's next

Nick's campaign is running. The dashboard is live. We're watching Nashville.

Meanwhile, we caught a podcast where the hosts were talking about wanting their show in multiple cities. Exactly the Advance use case. A show that big probably has agencies already, but the pitch is ready.

One customer live. One waiting for April. One cold outreach in the wild.