TechHauz rebuilt its infrastructure and cut fraud losses by 89%.
A digital gift card store running on WooCommerce where fraud was a daily threat. We migrated the server from CentOS 7 to AlmaLinux 8, enabled HPOS, restored Googlebot crawlability, and built an AI-powered fraud review pipeline on the Claude API.
// §01
The brief.
TechHauz sells digital gift cards instantly. That means every fraudulent order is a 100% loss, no chargeback recovery possible. The legacy fraud review process was manual, the server was on a deprecated OS, and Googlebot had stopped crawling product pages because of a Searchanise pagination bug.
Three problems, one engagement, all internal. We migrated the server, fixed the crawler issue, enabled HPOS in WooCommerce, and built TFR, an AI fraud review plugin that calls the Claude API on every flagged order and returns a structured risk verdict before fulfillment runs.
$ cat ./constraints.txt
· Server migration with zero downtime during the holiday season
· Googlebot must regain access to all product pages
· Fraud review pipeline must produce a verdict in under 8 seconds
· Stripe Radar AVS exception rules preserved exactly
· All automation must respect the locked structured-review template
// §02
The approach.
Four phases. Internal project, runs alongside agency client work. TFR plugin built in parallel with infrastructure migration.
01
Server migration
CentOS 7 to AlmaLinux 8 via cPanel ELevate. Staged in a snapshot, validated against production traffic patterns, cutover on a Wednesday with no holiday-period orders affected. PHP-FPM pool tuned post-migration.
02
Crawlability fix
Searchanise pagination was injecting client-side-only links. We patched the JS to render server-side fallbacks. Googlebot regained access within 11 days, indexed pages climbed back to baseline in 6 weeks.
03
TFR (TechHauz Fraud Review)
WooCommerce plugin that hooks the order-paid event, gathers Stripe Radar signals, geo + device data, and customer history, then calls the Claude API with a structured prompt. Returns a green/orange/red verdict in under 6 seconds average.
04
Ops hardening
Redis object cache, MySQL buffer pool tuned, HPOS enabled in WooCommerce, customer history statistics restored. Dispute counter-draft workflow templated as a PDF generator.
// §03
The stack.
// §04
The numbers.
Measured over the first 90 days post-deployment vs. the trailing 90 days. Fraud losses calculated from Stripe Radar reports and manual review escalations.
FRAUD LOSSES (90D)
-89%
absolute USD reduction
SERVER P75 RESPONSE
-62%
780ms → 295ms
INDEXED PAGES
100%
fully restored, 6 weeks
TFR REVIEW TIME
5.8s avg
from ~4 min manual
DISPUTE COUNTER WIN RATE
94%
from ~60%
REDIS CACHE HIT RATE
97%
from ~71%
“
TFR catches fraud my manual review used to miss, in seconds instead of minutes. The dispute templates win at 94% now. This is the infrastructure I wish I'd had three years ago.
— FOUNDER, TECHHAUZ
// related/
More e-commerce commits.
Or see the full ship log →
Got a WooCommerce store with fraud or infra problems?
Server migrations and fraud automation are scoped in 5-7 days and shipped in 4-8 weeks.