An MVP is not Phase 1 of the product. It is the smallest thing that tests our single riskiest assumption. Everything else we fake, do by hand, or defer until it earns its place.
The question this MVP answers: Will a real local business owner, cold and unsolicited, scan a mailed QR code, see an AI generated website of their own business, and want to claim and publish it?
If yes, the whole model has a foundation. If no, nothing downstream matters.
A single thin thread of the full funnel, run on a small hand picked batch of roughly 25 to 50 real local businesses.
Hand pick targets → generate a great site → postcard and QR → owner scans → sees their site → claims it → edits one thing → publishes live.
For version one we do not build the automated machine. We do the things that do not scale, and we build for real only the parts that are actually under test.
| Build for real (this is the test) | Fake or manual for now |
|---|---|
| AI website generation quality, the wow moment | Automated business discovery. We pick the list by hand. |
| The QR to preview to claim to publish experience | The dashboard and CRM. A spreadsheet is fine. |
| Funnel instrumentation: scan, view, claim, publish | Business Snapshot and Digital Health Score screens. Deferred, which matches the Notion MVP. |
| A minimal edit one thing path | Billing automation. We measure intent to pay and hand hold the first charges. |
Straight from the Notion Out of Scope list: SEO, Local SEO, AI improvements, reviews, booking, AI receptionist, promotions, lead capture, and the full dashboard. All real value. None of it helps answer the one question.
On a real batch we watch the funnel:
Both of these lean toward your side of the table, strategy and product direction, so I want your call on them before we build them in.
Our agreement puts me on execution risk: product development and technical direction. Here is what that concretely buys the company on this MVP.
Net: Paul funds and steers. I get us to real market signal with minimal spend, and I own that it ships and works.
Every sprint removes the single biggest obstacle to the scan, claim, publish journey. Right now that obstacle sits upstream of all the plumbing:
We do not yet know if we can generate a website good enough that an owner would actually want it.
So Sprint 1 attacks exactly that. It is the cheapest and highest information thing we can do.
Sprint goal: Generate 3 genuinely impressive AI websites for 3 real, hand picked local businesses. Each live at a shareable preview URL. Each good enough that we would bet an owner says yes.
Three sites live at URLs, each passing the quality bar, reviewed together by Michael and Paul.
Open the 3 real sites on a phone. That is the whole show.
Is the generated quality good enough to build the funnel around?
| Sprint | Focus | Output |
|---|---|---|
| 1 | Generation wow | 3 real sites live, quality bar met |
| 2 | Claim and publish | Owner can claim, edit one thing, go live |
| 3 | Postcard, QR, landing, tracking | A real mailable artifact plus funnel instrumentation |
| 4 | Mail the real 25 to 50 batch | First cold owner behavior data, go or no go on the hypothesis |
Roughly 4 weeks to first real market signal.
You flagged scope drift as one of our enemies, and you are right. The whole value of a one week sprint is the fixed box that forces us to prioritize, and the whole value of an MVP is fast learning. Drift breaks both: the deadline slips and the answer we need moves further away. Here is how we hold the line, every sprint.
The first place this bites is our own plan. The Phase 1 deliverables list already includes an automated discovery pipeline and a dashboard that the core question does not need yet. Hand picking the first batch and using a spreadsheet keeps Phase 1 pointed at the only thing we are trying to prove. That is the flag I want us to plant together.