Black Friday load testing: how to make sure your store stays online

Set your targets

Take last year’s busiest hour from analytics — sessions per minute, orders per minute and page views per session — then plan for 2–3× that to cover growth and marketing spikes.

Build realistic scenarios

• Browsers: homepage → category → filters → product
• Searchers: site search and layered navigation
• Buyers: add to basket → checkout → payment (use a test gateway)
• Logged-in customers with accounts and wishlists
• Background: bots, APIs and integrations

Tools

k6, JMeter, Gatling, Locust and cloud services like BlazeMeter or Grafana Cloud k6 can all generate peak load. Run tests from outside your network against a production-like environment, and monitor with New Relic, Datadog or CloudWatch while they run.

What usually breaks — and the fix

• Database: slow queries on filters and search — add indexes, use OpenSearch/Elasticsearch
• Cache: full-page cache bypassed by cookies — fix hole-punching
• Sessions: file sessions locking — move to Redis
• Autoscaling: scales too slowly — pre-warm capacity for peak
• Third parties: slow scripts and APIs — add timeouts and async loading

Run a war room

On peak days, have engineers watching live dashboards with a clear runbook: who can scale, roll back, switch off heavy features or enable a queue page. Agree it before the day.

Common mistakes to avoid

Frequently asked questions