How Lazada Cut 250ms from App Startup: Mobile Performance Secrets

With a population of 680 million, Southeast Asia’s e‑commerce market is booming, and Lazada has become the region’s leading platform, operating in six countries and supporting over 800,000 brands during the D11 shopping festival.

Startup Task Orchestration & Lazy Loading

The app’s launch is the first user‑facing step, so Lazada’s wireless team optimized task scheduling by grouping parallelizable tasks, reducing main‑thread occupancy, and minimizing lock contention. Key actions included splitting long‑running tasks, dispersing loadLibrary‑locked tasks, and deferring non‑essential work to after interaction.

Split lengthy tasks to shorten group duration.

Distribute loadLibrary‑locked tasks while preserving dependencies.

Defer unnecessary startup work until after the UI is interactive.

Startup Page Merging

By merging the splash Activity and the main Activity into a single Activity and switching views internally, Lazada eliminated one Activity creation and associated Binder calls, saving roughly 250 ms of startup time.

Full‑Link Dynamic Pre‑Request

Targeting low‑end devices and weak networks, Lazada introduced a full‑link prefetch framework that performs navigation preloading, idle‑time preloading, and dynamic routing without modifying existing APIs, accelerating PDP, Cart, and checkout flows by up to 450 ms.

Navigation preloading saves 100‑200 ms per page transition.

Idle‑time preloading predicts next pages and loads resources during idle periods.

Cart Asynchronous Incremental Update

Instead of full‑page updates, the server now returns only changed elements, and the client merges them with cached data, reducing request time and data volume, cutting Cart total‑time from ~1200 ms to ~750 ms (‑31.8%).

Cart Refresh Frequency Reduction

Refreshes on returning to the cart were limited to essential cases, decreasing refresh requests by 55.45% in the return‑to‑cart scenario and 35.28% overall, eliminating millions of redundant requests daily.

Checkout First‑Screen Pre‑Judgment

By reusing data from previous pages (Cart, PDP) and caching static elements, Lazada pre‑renders the checkout first screen, saving roughly 125 ms (cart‑to‑checkout) to 390 ms (PDP‑to‑checkout), with a total improvement of 528 ms when both paths are optimized.

Checkout Merge Request

A combined server endpoint encapsulating both cart submission and order rendering was introduced, moving the request trigger to the checkout page and reducing the checkout path latency from ~400 ms to ~64 ms (‑85%).

Pre‑Request & Early Rendering for the Cashier

Network calls were shifted earlier in the routing layer, and layout inflation was parallelized, yielding a 100 ms performance boost (701 ms → 605 ms) for the checkout UI.

Summary & Outlook

Overall, Lazada’s mobile performance initiatives delivered up to 23% faster Android launches, 20% faster iOS launches, and 34‑37% reductions in full‑link rendering times, while cutting redundant network traffic and preparing the platform for continued optimization.