Own the customer experience, rent the distribution.

Which Door Do You Actually Build?

By now the honest picture is uncomfortable, and worth sitting with rather than rushing past.

There are five different doors, they don't agree on the rules, at least two of the biggest ones run on a ranking process nobody outside the company can see, and one of them isn't really a door at all — it's a room you might already be standing in.

None of that is a reason to do nothing. It's a reason to stop asking "how do we get found" and start asking a better question: which of these doors is actually worth our effort, given where we already stand?

The proof that this isn't hypothetical: Walmart

Walmart didn't pick one door. It built its own — Sparky, a first-party shopping agent launched in mid-2025 — and also plugged into ChatGPT, and is rolling out to Gemini. Deliberately, not by accident: Walmart's own data showed conversion running roughly three times higher on its own site than inside a third-party chat window, so checkout stays home while discovery reaches outward.

Own the customer experience, rent the distribution.

That's not a theory. That's a top-three US retailer's live operating strategy, and it's the clearest answer available to the question this whole series has been building toward.

But "own the door" isn't the right answer for everyone, on day one

Here's the trap worth naming: reading the above and concluding we should build our own agent is exactly the mistake "Ambitious but Unready" retailers make. An agent embedded on top of an incomplete, inconsistent catalog just makes the gaps conversational instead of invisible — now the agent itself recommends a competitor, or worse, hallucinates an answer about your own product because the underlying data never told it otherwise.

So the real answer depends on where you're actually standing, on two separate questions: how ready is your data and infrastructure, today — not aspirationally, actually? And how much do you actually want to own the relationship, versus simply not losing ground?

Invisible — low readiness, low ambition. Most retailers, right now, and there's no shame in it — the category is a year old. Right move: a scoped audit, not a program.
Compliant but Passive — the data is actually fine, but nothing's been switched on. Often the fastest, cheapest win on the map: enabling what's already available rather than building anything new.
Ambitious but Unready — wants a Sparky of its own, but the catalog isn't there yet. Right sequence: readiness first, agent second, never the reverse.
Own the Door — Walmart's quadrant. Data's solid, appetite is real. This is where a first-party agent, built on your own platform, actually pays off.

THE FACTOR THAT DOESN'T FIT THE MAP

Amazon dependency doesn't sit on either axis, because it isn't about readiness or ambition — it's about how much of the customer relationship is already gone, structurally, regardless of what you do next. A retailer doing a third of its volume through Amazon has a materially different, harder conversation ahead of it than one doing five percent, no matter which quadrant the rest of the map puts them in.

Where this actually lands

The four-stage framework from earlier in this series, the schema.org and feed mechanics, the UCP profile — none of that was wrong. It's necessary groundwork for at least two of the five doors, maybe three. What it was missing was the map showing which door that groundwork actually opens, and whether that's the door worth spending the budget on at all.

That's not an article anymore. That's a conversation, and it starts with a genuinely honest answer to where you stand today — not where the last vendor's pitch deck assumed you'd want to be.

REFERENCES