E-Commerce Rescue: AI-Code Cleanup, 3-Min Loads Fixed

Client: Hello Gravel, Inc. — WooCommerce / WP Engine, via Contra Role: Rescue lead — audit, refactor, team + process build, feature delivery Timeline: 2 months core rescue · ~1 year overall (ongoing feature work) Stack: WordPress, WooCommerce, PHP, WP Engine, Cloudflare, Git, CI/CD, PR review

The problem. Pages took ~3 minutes to load. Under the hood: plugin hell, code pasted in from ChatGPT, cheap contractors copy-pasting the same function across the codebase, and deploys happening over FTP with no version control or history. The store technically worked and bled customers in practice — and nobody could change anything safely.

What I did. Audited the codebase end-to-end and untangled it — killed the duplicate logic, cut the plugin bloat, fixed the performance killers, then layered proper Cloudflare caching on top. Then I fixed the thing underneath the thing: stood up a real development process — version control, a dev team running pull-request reviews, and a staging → production flow — so shipping stopped being a gamble. From there, built revenue features directly: hyper-local product pages, a hardened order pipeline, and one-tap checkout.

Load time: ~3 minutes → 2–5 seconds, then a Cloudflare caching layer took it to milliseconds.

Hyper-local pages: region-specific product pages launched against a goal of 1 sale/day — they hit 2–3 sales/day immediately on rollout.

Order pipeline: flaky automation was dropping ~10% of orders (delays, lost revenue); after the fix, not a single order missed.

One-tap checkout: the system was so tangled that simply enabling Apple Pay / Google Pay crashed checkout — untangled and shipped properly, and ~15% of purchases moved to one-tap pay (sales that were leaking before).

A version-controlled, PR-reviewed codebase the team could finally ship on — no more FTP roulette.