Retail · 40,000 SKUs
Meridian Supply
Four times the usual Black Friday traffic, and not one checkout timeout.
38ms
cached time to first byte, from 214ms
4.1×
peak traffic held without queueing
+19%
checkout completion rate
9
days from audit to live cutover
The problem
Meridian sells industrial parts, which means deep filters over a 40,000-SKU catalogue and a customer base that shops from a phone in a van. On shared hosting, category pages took over two seconds to answer and the count queries behind every filter widget ran on each request. During campaigns the database queued, and the queue reached checkout.
The team had already doubled their plan twice. Both times the improvement lasted about a month.
What we did
1
Ran a read-only week on the live store. The slow query log put three query shapes above everything else, all of them filter counts.
2
Indexed the meta lookups and moved filter counts into a materialised table refreshed every ten minutes.
3
Provisioned three regions — Frankfurt primary with read replicas in Virginia and Singapore — and put full-page cache in front of everything except cart, account and checkout.
4
Added a warm-up crawl of the top 500 URLs, scheduled fifteen minutes before every campaign send.
“The first Friday after the move, I kept refreshing the monitoring because I did not believe it was that quiet.”
Ralph Edwards, E-commerce Director
Where it landed
Ninety days on, cached pages answer in 38ms and the logged-in cart sits at 340ms. The November peak came in at 4.1 times a normal Friday and the database never queued. Checkout completion is up 19%, which Meridian attributes mostly to the mobile catalogue finally feeling instant.
They are now on Pro with a named reliability engineer, and the quarterly audit has moved on to the parts of the stack that are not hosting.
At a glance
Sector
Industrial retail
Scale
40,000 SKUs, 1.2M monthly visits
Regions
Frankfurt, Virginia, Singapore
Plan
Pro, with migration
Timeline
Nine days, zero planned downtime