Six new micro-SaaS sites and one day to take them from "the site loads" to "a stranger can find it, pay for it, get email from it, cancel it and get help." The sites themselves already existed. This note is about launch day, which is where most small products stall.
We didn't improvise it. We ran one checklist with ten phases, the same one every time, and tracked it on one board: a row per item, a column per site, and a cell for each that reads done, partly done, not done or waiting on you.
What ran on its own
DNS and email: inbound addresses, SPF and DMARC, and sending-domain records for the email provider and the newsletter tool, all added by API.
Payments: one price per site, checkout tagged by brand, and a test purchase run on every site with one-use 100%-off codes: buy, access granted, cancel, access removed, then the test data deleted. No real card was involved.
Search: sitemaps, Search Console, Bing (verified by DNS, no sign-in) and IndexNow.
Analytics and newsletter: key events, source and segment tags, sync jobs and welcome sequences.
Support: a facts file per site and a support agent that drafts replies but doesn't send them yet.
QA: an hourly smoke test per site, customer emails rendered and reviewed before they could send, and a parity check against our most complete older product.
What needed a person
Every slow item was a human step. The email provider's free plan had a cap on sending domains, and we hit it with the second site. The fix was sign-ins to more accounts, which took most of a morning. Social accounts needed phone codes, and the cheap prepaid numbers we'd bought didn't receive them. One scheduler connection locked for days after too many code attempts. Linking a new Facebook Page to Instagram was refused ("isn't available right now") until a passkey approval went through later. And all of the social setup ran one screen at a time in a single shared browser.
The machine work ran unattended. Plan your human time first.
What went wrong
A newsletter filter matched any tag instead of all tags, and one broadcast went to the wrong list. The filter is now explicit, and the recipient count is checked before every send.
During cleanup we deleted the wrong DNS verification record. We put it back within minutes. Now we diff the records before deleting anything.
A live checkout page was left open in a browser with a real saved card. Nobody clicked, and it was canceled. Test purchases now run through the API with 100%-off codes only.
The board went stale: it showed some cells as not done after they were done. Every "done" now gets re-checked live before we call a site launched.
Where it stands
Honest scorecard at the end of the day: checkout, cancel, email, search and analytics are done on all six. Billing emails such as failed-payment reminders are built but switched off until review. Social seats are partly done: some sites have every account and some are waiting on phone codes. The support agent drafts and doesn't send yet. None of the six has a paying customer yet.
The checklist
We packaged the list as a pack. Every item says why it matters, what your AI can do alone, where you have to step in, and how you know it's done. There's also a Phase 0 we added after this day: one sitting before launch to check plan limits, test that your phone numbers receive codes, and decide who signs in where.
The Launch Checklist pack: the checklist, a launch board template your AI fills in, a support facts template and a made-up sample product to watch it work. $19. See the pack
These are our results, not typical, and the pack is a starting brief for your AI, not a promise of customers.
