Preventing Black Friday Failures with Load Testing for an Online Retailer

The Challenge

The client operated a mid-size UK e-commerce platform expecting 4 to 5 times their usual traffic during an upcoming promotional campaign. The previous year, checkout had slowed significantly during peak hours, but the cause was never identified. The team needed confidence that their updated platform could handle the traffic spike, and evidence of where any remaining weaknesses were, before the campaign went live. No existing performance test suite existed.

Our Approach

We designed a load testing suite using k6 covering the critical user journeys: browsing product listings, product detail page views, add to cart, the full checkout flow, and account authentication. Load scenarios were built to simulate realistic user behaviour, including think times and session abandonment rates derived from analytics data. We ran four scenarios: a baseline test at current traffic levels, a load test at 3x, a spike test simulating a sudden 5x surge, and a 4-hour soak test at steady elevated load. Results were streamed to Grafana for real-time analysis and threshold assertions were set to flag any degradation automatically.

The Outcome

The baseline and 3x load tests passed within acceptable thresholds. The spike test revealed checkout failures caused by database connection pool exhaustion above 400 concurrent users: the pool was configured for a fraction of peak demand. The soak test identified a memory leak in the session management layer that would have caused gradual performance degradation over hours of sustained elevated traffic. Both issues were remediated before the campaign launched. The promotional period ran without incident and the checkout conversion rate held steady at peak traffic.

Tools & Techniques