Layers

Product discovery isn't one path.

It's a funnel with two front doors: search and navigation.

Baymard finds roughly a 50/50 split on broad catalogs.

Obsess over one door and half your shoppers get the worse store.

Search relevance isn't one thing. It's four, stacked:

1. Textual: does it match the words?
2. Business: do you want it ranked?
3. Behavioral: what do shoppers actually pick?
4. Semantic: does it match the meaning?

Tune one, ignore the rest, and search feels "off."

For a decade, search was table stakes: a box that returned rows.

Then AI raised the ceiling, and most stores didn't move.

Baymard: 56% still fail basic search UX.

When the majority is bad at something shoppers do every day, being good at it is a moat again.

Relevance is not the goal. It's the ticket to play.

A perfectly relevant result set that's all low-margin, out-of-stock, or off-strategy still loses.

Relevance gets you in the room. Revenue wins it.

Where AI is mostly hype: the promise that adopting it automatically prints money.

McKinsey's 2025 read:

Most organizations now use AI somewhere.

Only about 39% report any real EBIT impact.

Adoption is easy. Captured value is the hard part.