Work / Callstack client, Fortune 500 company

Profiling and speeding up a Fortune 500 B2B ordering app

Performance and architecture audit that became the client's migration plan, then the fixes: splash render 18x faster, basket re-render down 89%.

Organisation
Callstack client, Fortune 500 company
My role
Lead engineer on the audit, then implementation
Period
Oct 2025 - Feb 2026

Context

A global B2B ordering platform, React Native on mobile plus a web app, built by a large multi-vendor team. Callstack was asked to find out why key screens were slow and what to do about it.

Problem

Key screens re-rendered far more than needed because of how Redux state and selectors were shaped. The client needed a plan it could act on, not a list of symptoms.

Approach

  • Led a performance and architecture audit of the app. The client approved it as its migration plan.
  • Moved Redux selectors and reducers to work with the React Compiler, so components stop re-rendering on unrelated state changes.
  • Started a gradual TypeScript migration and added pre-commit TypeScript and ESLint checks.
  • Wrote scripts to measure memory use, so later changes can be compared against a baseline.
  • Automated over-the-air release steps.

Key decisions

Measure every change before and after
Each pull request carries React Profiler numbers for the screen it touches. A claim without a before number is not a result.
Report per screen, not one "app speed" number
A single headline figure hides where the time goes. Per-screen numbers tell the team what to fix next.
Migrate to TypeScript gradually
A big-bang migration on a codebase shared by many vendors would block everyone. File-by-file conversion with pre-commit checks keeps it moving.

Outcomes

  • Reworked splash-screen state and rendering: total render time went from 131.59 ms to 7.24 ms (18x faster).Source: Pull request, React Profiler
  • Cut basket re-render time after carousel updates from about 1,800 ms to about 200 ms (89% less), with dispatches down from 7 to 2.Source: Pull request, React Profiler
  • Cut a nested cart-footer update from 257.82 ms to 46.09 ms (82% less).Source: Pull request, React Profiler
  • Made the rewards screen about 3x faster and the filter sheet 27% faster.Source: Pull request, React Profiler

The client is not named. Timings are React Profiler measurements on development builds, per screen. They show relative improvement, not end-to-end app start time.

Stack

  • React Native
  • Redux Toolkit
  • Reselect
  • React Compiler
  • TypeScript
  • MMKV
  • React Profiler
  • SonarQube