George Kelly / Research & practice
Agentic commerce.
An operator’s field guide.
My working notes on commerce economics, product data, and practical AI. Each guide takes a question through the source material, assumptions, and a method you can examine.
Explore the guidesThe working library
Read it. Put it to work.
36 connected guides.
Clear definitions, worked examples, and useful artifacts.
Explore a collection
AI search measurement ↗
From AI impressions to actual visits: six connected resources, a metric dictionary, review worksheets, and a click-planning calculator.
Agentic commerce starts with an operating model.
A practical framework for choosing agent-assisted commerce work, setting decision rights, and proving that the result is worth the cost.
ROAS calculator: break-even and contribution
Calculate ROAS, contribution after advertising, and the break-even return for your margin. Includes worked examples and a margin comparison.
Product data an agent can use without guessing.
A practical product-data contract for variants, dimensions, availability, provenance, and unknown values, with a downloadable JSON example.
From request to release: a supervised commerce workflow.
An original seven-stage operating checklist for turning agent-assisted commerce work into an inspectable, authorized, verified release.
Measure AI visibility without confusing it with traffic.
A measurement framework that separates crawler access, answer citations, referral visits, and qualified commercial outcomes.
Make a website easy for agents to find, read, and use.
An evidence-based guide to crawler access, HTML, structured data, Markdown, llms.txt, feeds, and real tool descriptions.
How to measure AI citations and visible brand attribution
A repeatable rubric for checking whether an AI answer consulted, linked, displayed, or named your source. Includes a blank ledger and calibration examples.
How do you check AI product copy before publishing?
A claim-by-claim review method, six failure cases, and a reusable evaluation brief for AI-assisted product descriptions.
Did AI save time after review and rework?
Compare the full cost of accepted work, including preparation, supervision, review, and corrections. Includes a local calculator and a reusable time ledger.
AI visibility metrics: impressions, citations, referrals, and outcomes
A source-linked metric dictionary for people and agents building AI visibility reports. Includes units, formulas, interpretation limits, and a versioned JSON download.
Content API 410 Gone: diagnose the sunset and review your migration
Identify a Content API sunset error, find the integration owner, and review a Merchant API cutover with eight evidence checks and a downloadable worksheet.
Traffic fell: repair, refresh, or publish a new page?
A practical editorial decision worksheet for separating technical faults, incomplete measurements, overlapping answers, and useful new research when traffic falls.
A working contribution API integration: 14 acceptance cases
Run a real, bounded contribution API check. Inspect exact requests and responses, preserve losses and null ratios, and hold mismatched currency or cohort metadata before calling.
How to find referral traffic in GA4, including AI
Find referring sites in GA4, separate Referral from AI Assistant, and check Google AI traffic. Includes report steps and a downloadable review worksheet.
Google product category vs. product type: choose the right mapping
Separate Google’s predefined taxonomy from your merchant product_type hierarchy. Review category IDs, breadcrumbs, ambiguous products, and six fictional mapping cases.
Can your stock support more ad spend?
An inventory and ad-spend calculator with daily stock checks, a delayed-delivery example, and a downloadable decision worksheet.
Missing GTIN: when to use identifier_exists in Google feeds
Decide whether a missing GTIN is unknown or genuinely unassigned. Use six worked cases and an evidence worksheet before changing identifier_exists.
Why feed price and Offer.price can differ without an error
Reconcile regular and sale prices across a feed, product page, checkout, and structured data. Keep currency and context explicit; download eight review cases.
Product compatibility data: prove the exact accessory match
Separate compatible products from related products. Use exact model pairs, revisions, adapters, and source evidence, with a downloadable decision ledger.
Why a product feed check can pass while the records disagree
Five reproducible synthetic source/feed pairs separate single-record checks from price agreement, exact identity and fresh evidence. Includes inputs, results and runner.
Product data ownership: resolve a price conflict
A field-ownership register for PIM, ERP, storefront, and feed data, with a worked correction and explicit release checks.
Comparing product dimensions across country pages: a 30-listing audit
Eight complete comparisons, twenty retrieval gaps and two axis gaps: a reproducible audit of displayed product dimensions on sampled IKEA US and GB pages.
Product dimensions: units, examples, and comparison tool
Understand product vs package dimensions. Compare two records in inches, cm, mm, or metres, see differences, and download the result. No signup or upload.
Product feed specification: Shopify, Google, and ChatGPT
Map 20 product feed fields across Shopify, Google Merchant Center, and ChatGPT. Compare IDs, price, stock, and variants; download the JSON reference.
ChatGPT product feeds: sale prices, stock changes, and stale-data checks
Check current price and stock evidence against a ChatGPT product-feed snapshot. Includes a local checker, six fictional cases, and explicit timestamp rules.
Can you copy a product feed unchanged? A 12-case test
Inspect twelve synthetic product records or check your own JSON locally. Review explicit mappings, unresolved fields, and supplied evidence; download reproducible results.
SKU vs GTIN: differences, barcodes, and variant IDs
Compare SKU, GTIN, UPC, and variant IDs with examples. Check a product record in your browser for identifier and variant issues; download the review.
Does the product page match the variant you submitted?
A six-case acceptance test for variant identity, price, availability, and missing facts, with reusable JSON fixtures.
Product variant URLs: what eight public pages actually selected
An executed eight-variant study of URL selection, rendered options and JSON-LD offers, with captured facts and replayable extraction tests.
Why retail-media revenue and settled orders disagree
Bridge ad-attributed sales to a defined order ledger, preserve unresolved differences, and keep bank payouts separate. Includes a worked USD example and CSV/JSON worksheet.
Is the campaign still profitable after returns?
A worked order-cohort reconciliation for refunds, recovered inventory, return costs, and ad spend, with a reusable review template.
Sale price effective dates: do they schedule the promotion?
Compare Google Merchant Center sale scheduling with OpenAI native updates and confirmed Google-compatible date metadata. Eight manual review cases and a reusable JSON record.
Why Search Console, GA4, and Vercel show different traffic numbers
Compare clicks, sessions, and page views without forcing the totals to match. Includes a worked example and a reusable reconciliation worksheet.
How to read Search Console’s generative AI report
Measure Google AI Overviews and AI Mode exposure, avoid double counting, and turn an impression report into a useful content decision. Includes a review worksheet.
Search impressions but no clicks: a diagnostic and planning calculator
Work backward from a click target, test CTR assumptions, and choose a useful response to search exposure with few visits. Includes a local calculator and worked scenarios.
Shipping and return policy data for shopping agents
Compare delivery charges, handling time, return windows, and policy exceptions in one scoped record. Includes fictional offers and a reusable JSON worksheet.
One number rarely tells the whole story
5× ROAS.
What’s left?
Revenue efficiency and commercial contribution answer different questions. Change the assumptions and see what remains after advertising.
Use the contribution calculator- Net revenue
- 1,000
- Product + variable costs
- −700
- Advertising
- −200
- Contribution after ads
- 100
Before fixed overhead. One currency throughout.
Built for people. Accessible to agents.
Good ideas should
travel well.
Every guide is available as HTML, Markdown, and JSON, with the same sources and scope notes. Subscribe to the feed or bring a useful framework into your next workflow.