We use cookies
To improve your experience. Cookie policy
RetailOS · E-Commerce Technology · 2024
RetailOS's storefront was bleeding conversions — a 6.2-second LCP on mobile was costing them an estimated $340K/month in lost revenue. We rebuilt the frontend from the ground up in Next.js and cut load time by 82%.
Key outcomes — E-Commerce Technology
1.1s
LCP (mobile)
from 6.2s
2.3%
Conversion rate
from 0.8%
+$180K
Monthly revenue
net new
−70%
Server instances
origin fleet
§ 01 / Problem
A React SPA built in 2019 was serving a product catalogue of 120,000 SKUs with no server-side rendering, unoptimised images, and a 4.2MB JavaScript bundle. Mobile conversion rate was 0.8% against an industry benchmark of 2.1%. Seasonal traffic spikes caused the origin servers to fall over despite running on AWS.
§ 02 / Approach
We ran three weeks of performance measurement before writing any production code — real user monitoring, waterfall analysis, and bundle profiling across 50 representative product pages. For each intervention we modelled the conversion impact per engineering day and executed in strict priority order. The migration was phased: highest-traffic product pages first, with conversion lift measured and confirmed before rolling out to the full 120,000 SKU catalogue.
§ 03 / Solution
We migrated to Next.js App Router with React Server Components, replacing the monolithic client bundle with component-level code splitting. Implemented Incremental Static Regeneration for product pages with a 60-second revalidation window. Image pipeline rebuilt with next/image, WebP conversion, and a CloudFront CDN. Checkout flow rebuilt with optimistic UI and a background cart API to eliminate perceived latency.
§ 04 / Result
LCP dropped from 6.2s to 1.1s. Mobile conversion rate rose from 0.8% to 2.3% within 30 days of launch — recovering the estimated lost revenue and adding a net $180K/month in incremental sales. The origin fleet scaled down from 40 instances to 12.
1.1s
LCP (mobile)
from 6.2s
2.3%
Conversion rate
from 0.8%
+$180K
Monthly revenue
net new
−70%
Server instances
origin fleet
Tech stack
We knew our site was slow, but we didn't fully understand the revenue impact until stackloader showed us the conversion data. Going from 6.2s to 1.1s load time added $180K in net-new monthly sales. The ROI conversation with our board was very short.