DEV3Technologies
All case studies

E-commerce Plus

Cutting checkout latency by 77% for an omnichannel retailer

A legacy monolith checkout rebuilt on Next.js with edge caching, idempotent payment orchestration and real-user monitoring — conversion rose from 2.1% to 3.4%.

Next.jsMongoDBRedisNode.jsPayment GatewaysRazorpayEdge CachingAWS
Industry
Commerce
Region
United Kingdom & India
Duration
14 weeks
Team
6 engineers
Services
Web DevelopmentCustom SaaS Engineering

Impact

Checkout latency (p75)

190 ms77%

Before: 820 ms

Conversion rate

3.4%62%

Before: 2.1%

Failed-order tickets / month

4288%

Before: 340

Live project preview

https://example.com
Open

The challenge

Checkout pages averaged 820 ms server response at the 75th percentile and spiked past 3 seconds during sales. Payment retries occasionally double-charged customers, support tickets about failed orders were rising every quarter, and the team had no reliable way to tell which release caused a regression.

Our solution

Edge-first rendering

We split the checkout into cacheable shells and personalised islands, served from the edge with stale-while-revalidate for catalogue and pricing data.

Idempotent payment orchestration

  • Every payment intent carries an idempotency key persisted in MongoDB before the gateway is called.
  • Razorpay and a secondary gateway sit behind a single adapter with automatic failover.
  • Webhooks and client callbacks share one fulfilment path, so either can complete an order safely.

Observability

Real-user monitoring and release markers made every Core Web Vitals change attributable to a specific deployment.

Results

Within six weeks of launch the new checkout handled the festive sale peak without a single timeout.

  • Checkout server response fell from 820 ms to 190 ms at p75.
  • Conversion rate improved from 2.1% to 3.4%.
  • Duplicate-charge incidents dropped to zero.

Related case studies