Commerce & websites · Client engagement
Shopify engagement for an education business
Storefront development, analytics, search visibility, and internal tools for an online education store. Engaged since February 2026, with weekly contracts since March.
What the engagement covers
Four layers of work around one store, from the pages customers see to the checks behind the product content.
- Role
- Development, analytics, and search
- Period
- February 2026 to present
- Status
- Ongoing, weekly contracts
- Platform
- Shopify, Cloudflare
- Built with
- Shopify Liquid
- Cloudflare Workers
- Cloudflare D1
- GA4 Data API
- Search Console API
- Resend
- SQLite FTS5
- 33
- findings in the search and AI-visibility audit, each with an owner and an effort estimate
- 8,364
- indexed reference passages, matched to printed page numbers at 99.1%
- 1,625
- commits on the engagement's main repository
The engagement
The client runs an online store that sells study materials. I work under weekly contracts that name each deliverable. The scope started with theme changes and grew into the store's analytics, its search visibility, and the tools used to produce its products.
The store runs on a forked Shopify theme behind Cloudflare, with Workers handling the automations Shopify does not do on its own.
Storefront
- Product page variants, reusable theme sections, badges, and navigation changes.
- Review import and storefront configuration.
- An email preference center with signed unsubscribe links.
Analytics the owner can read
A scheduled Worker pulls GA4 and Search Console data into a D1 database, and a client dashboard reads from that database. The owner sees traffic, search clicks, and impressions without opening either Google product.
Search and AI visibility
The June 2026 audit covered classic search, answer engines, and AI crawlers: 33 findings, each with an owner and an effort estimate. Structured data was parsed by a program instead of read by eye.
A reference library for fact-checking
Product content is checked against a library of reference books. I indexed them into 8,364 searchable passages and matched each passage to the page number printed in the book, at 99.1% accuracy.
One book had 1,478 pages whose embedded page labels disagreed with the printed numbers, by as many as 168 pages. That mismatch needed a separate tool that reads the printed page number from each page image.
Scope & limitations
The client is not named. The store's traffic, revenue, source code, dashboards, and brand assets are not shown.