Who this is for
We work with four ecommerce shapes. The common thread is deliberation: enough price or complexity that a buyer researches before purchasing.
DTC & consumer brands
You own the storefront and the customer relationship. The brand entity is yours to control, and the constraint is usually that the product entity underneath it was never structured at all.
Marketplace sellers
Amazon, Walmart, Target+. Most of what an assistant knows about your product comes from a surface you do not own, which changes the work from publishing to correcting.
Subscription & replenishment
Recurring purchase, so the buying question is asked once and the answer compounds. Being the recommended option at the moment of first choice is worth more here than in any other model.
Considered-purchase specialty retail
High-ticket, research-heavy, long deliberation, antiques, furniture, instruments, equipment. The qualifying shape: buyers research for weeks and ask assistants direct comparison questions before they ever reach a store.
Most of what AI says about your products is not on your site
For commercial queries, roughly two thirds of the retrievable web an assistant draws on is content the brand does not own, review platforms, category-specialist comparison sites, marketplace listings, roundups. In some categories that rises past 80%. This is the single biggest structural difference between ecommerce and a service business, and it means a strategy that consists only of improving your own pages is working on the minority of the surface.
You have two entity layers, not one
A service business has one entity to stabilise: the business. An ecommerce catalog has the brand and the product. The SKU is its own entity, with its own identifiers, its own representation across marketplaces, and its own ways of being described inconsistently. No service vertical has this layer, and it is where most of the new work sits.
The measurement protocol has to change shape
We publish an X/12 sampling protocol and run it across every other vertical we work in. It does not transfer to a catalog unmodified, and we would rather say so than sell you a number that does not mean anything. What we run instead is described below.
What we run instead of a flat store-wide score
We publish our measurement protocol and run it across every vertical we work in. For catalogs it is restructured rather than reused, and here is exactly how.
Not one flat score for the whole store
A single twelve-question run cannot represent a catalog. We run X/12 per product-category cluster, twelve questions for each of your top two to four SKU categories, because categories differ so sharply in how answers get built that averaging across them produces a number with no meaning.
Questions weighted by how third-party the category is
In a category where 80%+ of the citation surface is off your domain, most of the twelve ask whether your product is correctly and favourably represented in the top roundups and review sites, not whether an assistant says your brand name. Direct brand mentions are structurally rare there no matter how good your site is, so scoring them would measure the category rather than the work.
Two tiers, reported separately
Brand-level mention, which keeps continuity with how we measure in every other vertical, and product/SKU-level surfacing, which is the layer that only exists here. They are never collapsed into one figure.
Single passes labelled as snapshots
Run-to-run instability is documented in this space, around one in ten surfaced domains differs between identical pulls minutes apart. Catalog breadth multiplies that. A single run is a snapshot and is labelled as one.
Search authority, qualified demand and AI Share of Answer are measured together, on one method, with every result reported and every completed action recorded in the Work Ledger.
Others give you a score. We publish the method that produced it, and the ledger of what was done. Read the X/12 Protocol.
Fair questions
Do you work with any ecommerce business?
No. The shape that fits is considered purchase, enough price or complexity that buyers research before they buy and ask assistants direct comparison questions. Impulse and commodity categories are a poor fit for this work and we will say so.
Is this different from your home-services work?
Substantially. Two of the six audit layers need rebuilding rather than adapting, the citation surfaces are mostly third-party rather than GBP and directories, and the measurement baseline is redesigned. The framework holds; the delivery differs.
What platforms do you work on?
Shopify, WooCommerce, BigCommerce and Magento. Platform matters more than it used to, crawler access and structured-data control differ meaningfully between them, so the audit includes a platform-aware pass rather than one robots.txt read.
What does it cost?
Inside the Visibility Engine, from $3,500/month. The diagnostic that comes first is $3,500–$7,500 and is credited against the first program invoice.
Start with a measurement, not a pitch
A free Visibility Check shows where your catalog actually stands across Google and AI answers, before any conversation about scope or price.
Get my free Visibility Check