Every 100ms of additional page load time costs measurable conversion rate. For e-commerce, the correlation is immediate and quantifiable. For SaaS, it shows up in churn and NPS. For enterprise software, it shows up in adoption rates and helpdesk tickets. Application performance is a business metric, and treating it as an engineering vanity metric is one of the most expensive mistakes a product organisation can make.
This guide covers the highest-leverage performance interventions across the full stack, in the order that produces the greatest return for the engineering investment.
Database: The Most Common Bottleneck
In most applications, the database is the source of the majority of response time. N+1 query problems, missing indices on foreign keys used in joins, unoptimised full-text searches, and lack of connection pooling are consistent culprits. Before optimising anywhere else, run your slowest endpoints through a query profiler and address the top five slow queries. The returns are almost always immediate and dramatic.
"We reviewed an application where 85% of response time came from 3 unindexed queries. Adding the indices took 20 minutes and cut p95 response time by 60%. This is the kind of win you find before you start talking about CDNs and caching strategies."
Caching: The Right Layers
Application-level caching (Redis or Memcached) for expensive, frequently-accessed, slowly-changing data is almost always the correct first caching investment. HTTP caching with proper Cache-Control headers is often overlooked on API responses that are actually cacheable. CDN caching for static assets is table stakes — if you are serving static files from your application server, you are leaving substantial performance on the table.
Frontend: The Perception of Speed Matters
Core Web Vitals — LCP, FID, and CLS — measure not just technical performance but perceived performance. Server-side rendering for initial page load, code splitting to eliminate unnecessary JavaScript, and optimised image delivery (WebP, AVIF, responsive sizes) are the highest-leverage frontend interventions. For React applications specifically, profiling with the React DevTools profiler often surfaces component re-renders that are consuming far more CPU than necessary.
Ready to build something remarkable?
Our engineering and AI teams help ambitious organisations design, build, and scale intelligent systems. Let's talk about your next challenge.