AI is already choosing products. Make sure it gets yours right.
Connect your catalogue and watch AI shopping agents try to serve realistic customer needs. Herm shows whether they find the right product, answer the questions that matter, trust the underlying evidence and send the shopper somewhere they can actually buy.
Product-feed URL, XML, CSV, JSON, Google Merchant Center or API. Tests repeat on a schedule: read the methodology.
- 01 Shopper need
- 02 Agent
- 03 Recommendation
- 04 Evaluation
“I’ve got rosacea-prone skin and I want to start using vitamin C. I reacted badly to one last year, so I need something that won’t sting. I’d like to stay under £40.”
Velora Bright Boost C
Your feed can pass every check and still fail the customer.
Feed validation asks whether your information is structurally accepted. It is a necessary check, and it is not the same question as the one a shopper is asking.
Product Readiness asks whether an AI shopping agent can use that information to satisfy a real customer's buying need correctly: the right product, the questions resolved, the claims supported, and a place to buy.
Four things have to be true for a sale to survive an agent.
Every shopping test is evaluated across the same four stages. A journey that clears three of them still ends in a customer who cannot buy the right thing.
Can the agent retrieve the right eligible product?
Does the available product information resolve the shopper's important questions?
Agent Answerability lives here.
Can the agent rely on the identity, claims and commercial facts it is using?
Can the shopper reach an appropriate place to buy?
Close covers reaching an appropriate place to buy: a valid product URL, a retailer destination, availability and the right geography. Herm does not process payment, create carts or execute checkout.
One number for the health of the catalogue. One for the customer's experience of it.
The Be Sellable Score is a diagnostic health view across Find, Answer, Trust and Close. It tells a product team where to work.
The Sellable Intent Rate is the customer-journey outcome: the share of tested buying situations that end with an appropriate, evidence-backed and commercially valid recommendation with a working path to purchase.
Both are reported per market, per category and over time. Scoring weights and thresholds are not published.
Answer is the weakest dimension and the score fell after a catalogue update. Section 06 shows what a single dimension drop resolves to.
See what your product information can answer, and what it cannot.
Answerability measures decision-relevant information, not field completeness. A record can be 100% complete against a feed spec and still leave the questions that matter in real buying decisions unresolved. Select any cell to see the evidence behind it.
Vitamin C Serums · “A gentle form of the active?”
UnknownThe agent cannot resolve this question. It will either omit the product or answer from a third-party source you do not control.
Correct according to the catalogue isn't always correct for the brand.
A recommendation can be factually consistent with your feed and still be commercially wrong: the advanced retinoid for the first-time user, the discontinued formulation over the current one, the unapproved claim.
Brand Intent lets you state what correct means. Herm then evaluates both factual accuracy and commercial intent.
Don't stop at “19 journeys failed.” Find the problem behind them.
A list of failures is not a work plan. Herm groups failing journeys into the product-information problem they share, tells you how many products carry it, and what to add.
Add authoritative form, fragrance and tolerance-study information.
Turn a failed journey into a better product record.
Each finding carries the products it affects and proposed product-data changes with their evidence. You approve what is accurate, export it, and apply it through the workflow you already use.
- 01 Finding 19 failed journeys, one root issue
- 02 Affected products 214 products, 34 variants first
- 03 Proposed data field-level changes, each with evidence
- 04 Review queued for brand review
- 05 Approve brand decides what is accurate
- 06 Download / export no write-back to your PIM
- 07 Source updated, re-test you publish, Herm re-runs the same journeys
What Herm does not do. Herm does not write into your PIM. Approved changes are exported and applied through your existing product-data workflow. Herm then re-tests the same journeys.
Fix it. Run the same shopper again.
Readiness is not an audit you pass once. The same buying situation, the same evaluation context, re-run after the product data changes.
This is measured improvement inside the Product Readiness evaluation. It is not a claim about external sales.
Product data updated: 34 variants, form and fragrance evidence approved 22 Aug, re-tested automatically.
Sellability changes when your catalogue does.
Herm tracks the environments and standards that decide whether your products are eligible to be sold by an agent. Products change, channels change, and a journey that passed last month stops passing. Herm tests on a schedule, watches for the changes that matter, and raises the ones worth acting on.
- 09 Aug Product updated 34 variants changed in the connected source
- 09 Aug Test re-run 50 journeys · vitamin C serums & active form
- 10 Aug Answerability decreased Vitamin C Serums · gentle-form questions ↓ 11
- 10 Aug Alert raised Sellable Intent Rate below threshold
- 22 Aug Product enriched form + fragrance evidence approved
- 22 Aug Journey passed journey 31 re-tested · same intent ↑ 9
Tracked, not certified. Herm follows the published requirements of these standards and environments and reflects the relevant ones in tests.
First be recommended. Then be sellable.
- 01 Brand Visibility Live Rung 01 · Be Visible Does AI consider and recommend the brand?
- 02 Product Readiness Live Rung 02 · Be Sellable Can AI identify and correctly sell the product behind that recommendation?
The brand enters consideration, but detailed customer requirements expose product-information gaps.
The catalogue would serve the customer well, but the brand is never in the conversation where that would matter.
Four different jobs. Only one of them is a shopping test.
Stores and manages product information.
Checks whether the feed meets structural and channel requirements.
Builds the consumer shopping experience.
Tests whether AI agents can correctly use the brand's product information to satisfy customer buying needs.
Common questions
Do you need our product feed?
Yes, for the full assessment. Product Readiness evaluates the commerce and product information a brand actually provides, so it needs a connected source: a product-feed URL, XML, CSV, JSON, Google Merchant Center or an API.
Is this a feed validator?
No. A validator asks whether your information is structurally accepted by a channel. Product Readiness asks whether an AI shopping agent can use it to satisfy a real customer's buying need. A catalogue can pass every structural check and still fail a shopper, which is the case the tests are built to surface.
Does Herm write changes into our PIM?
No. Proposed product-data changes are shown with their evidence for you to approve, and approved changes are exported for your team to publish through the systems you already own. Once the source is updated and re-connected, Herm re-runs the same journeys.
Does a shopping test complete a purchase?
No. Close covers reaching an appropriate place to buy: a valid product URL, a retailer destination, availability and the right geography. Herm does not process payment, create carts or execute checkout.
Does Herm certify compliance with ACP or UCP?
No. Those are ecosystem specifications Herm tracks. Tracking a standard does not mean Herm certifies compliance with it or performs official validation on its behalf. Where Herm has a specific readiness test for a requirement, the result says so precisely.
Are the scoring weights published?
No. The concepts, units and boundaries are public and documented in the methodology. The evaluator, the scoring formula, the weights and the thresholds are not.
See what happens when AI shops your catalogue.
Connect a feed, pick a category, and watch a shopping agent try to serve fifty realistic buying situations.
Sample figures throughout.