The challenge
A top-100 global retailer had spent two years quietly outgrowing its existing CDN deployment. With a major Black Friday campaign 90 days out, the team flagged two problems they couldn’t solve internally:
- Catalog cache hit ratio was 71% — and origin egress was scaling alarmingly with traffic
- Scraper traffic from competitors had reached an estimated 18% of total catalog requests
- Previous peak seasons had required emergency origin scaling and triggered Tier-2 incidents
The platform team needed a 90-day path to confident peak readiness.
The approach
We split the work into three concurrent tracks:
Cache architecture
A full audit of cache key construction, Vary headers, cookie handling, and edge personalization logic surfaced eight specific opportunities. We redesigned cache behavior for the catalog and PDP routes, lifting the catalog hit ratio from 71% to 94% over six weeks.
Bot management
Akamai Bot Manager was deployed in observation mode for two weeks to establish baselines, then progressively moved to challenge and block. Custom rules targeted the specific scraper patterns identified in the audit. Legitimate partner integrations (price comparison, affiliate networks) got explicit allowlists.
Load and chaos testing
Six weeks of progressively aggressive load tests validated the new architecture — including a final test at 100x baseline that uncovered a checkout race condition the team had never previously been able to reproduce. Fixed before launch.
The results
Black Friday peak traffic hit 12x baseline. Origin auto-scaling never triggered. No incidents above P3.
Conversion rate during peak campaigns lifted 22% versus the prior year, driven primarily by page performance improvements (LCP improved 31% at the P75). Catalog cache hit ratio settled at 94%, taking origin egress costs down 58% versus the run-rate from before the engagement.
The bot defense work blocked an estimated 14% of total request volume as confirmed scrapers — without measurable impact on legitimate users.
What made it work
The platform team was clear about priorities from the start: peak readiness was non-negotiable, but the cache work also had to deliver year-round value. By tracking both peak metrics and steady-state economics, we kept the engagement honest about what was actually working.
The willingness to do aggressive load testing — including breaking things in pre-production — paid off enormously. The single race condition surfaced by the 100x test would have caused a major outage during peak.