Building something that isn't a tenant

I'm building Internet Research Unit. It's alternative data for ecommerce — revenue, SKUs, tech stacks, and consumer trends across millions of DTC brands, for brand owners, agencies, app developers, and the people who finance them.

The reason I'm building this one is that it's the first thing I've ever built that isn't a tenant.

The pattern I'm trying to break#

MySpace, where my first audience lived. Facebook's organic reach, which my stores depended on. The ad auctions. Exposely's whole supply side, priced by reach algorithms nobody outside those companies could see. Then four companies inside Shopify's ecosystem, one of which died because a file was deprecated.

In every case I made plans that assumed the current arrangement would hold, and in every case it didn't, and I've written the same lesson down about five times now without changing my behavior.

I want to be precise about what I'm reacting to, because "don't build on platforms" is the wrong lesson and I don't hold it. Shopify is the best distribution I've ever had access to. I built four companies there and two of them worked. Refusing that on principle would have cost me more than platform risk ever did.

The thing I'm reacting to is narrower: I've never once built something where the data I depend on comes to me by right rather than by permission.

What's different about this one#

IRU's input is the public surface of the ecommerce industry. Storefronts, what they sell, what they run on, how it changes over time. That's observable from outside, across the whole market, without any single company deciding whether I get to see it.

That's a genuinely different structural position than anything I've built. There's no API key that gets revoked. There's no ranking algorithm that decides my growth rate. There's no one deprecation notice that ends the company.

I'm aware this is not the same as invulnerable. There are things that could go wrong here that I can see and probably several I can't. But the failure modes are diffuse rather than concentrated in one counterparty, and after Vigilance I want diffuse.

What I'm giving up for it#

The distribution, and it's a serious cost.

The Shopify App Store gave me a qualified customer who searched for their problem, found me, and installed in a click with billing handled. Every company I built there started with that. IRU has none of it. I'm back to explaining what the thing is to people who weren't looking for it, which is the exact painful first act I went through at Exposely for three years.

I know what that costs because I've paid it. I'm choosing to pay it again on purpose, which is either maturity or a very expensive reaction to one bad outcome.

The honest uncertainty#

There's a version of this where I've overcorrected. Vigilance was one company that died from one specific exposure — a product that was a platform behavior rather than one that used a platform. I already knew how to distinguish those; I wrote it down in 2018. The correct response might have been "apply the test you already had," not "restructure my whole approach to what I build."

Instead I've started something with much worse distribution because I want to own the ground. That's a real trade and I've made it emotionally as much as analytically. I'd rather say that than pretend it was pure strategy.

Ask me in three years whether owning the input was worth losing the app store.