How Gaode Map Halved Core Link Latency: A 3‑Year Mobile Performance Overhaul

Overview

Since 2019 Gaode Map has been optimizing the performance of its entire request chain. After three years of continuous improvement the core‑link latency was reduced by 50%, delivering a noticeable boost in user experience.

Three‑Step Approach

The overall methodology consists of three parts: clearly defining performance bottlenecks, reverse‑order specialized problem solving, and forward‑order long‑term control.

Define Bottlenecks Accurate measurement standards and quantifiable targets are essential. Metrics must reflect real user perception and guide focused, efficient execution.

Reverse‑Order Specialized Solving Performance issues often span multiple product‑research‑test teams. Starting from the problem, teams work backwards in a dedicated project to rally resources, set clear goals, and achieve quick results.

Forward‑Order Long‑Term Control After fixing a problem, continuous forward control prevents regression. Ongoing monitoring and systematic governance keep performance gains stable.

Define Standards

The first step is to establish clear standards for measuring the first‑screen load time, which heavily influences user perception. Different device tiers (high, medium, low) are identified based on hardware scores, and the first‑screen definition varies by business scenario, product usage frequency, and engineering instrumentation.

Identify Performance Bottlenecks

Given Gaode’s extensive codebase, manual analysis is impractical. The team uses tools and methodologies to quickly locate time‑consuming points across three dimensions:

Device Dimension Analyze performance on low‑end devices; issues that appear on cheap phones may be invisible on high‑end models.

Business Parallel Dimension Prioritize scenarios (e.g., search page) that receive the most user traffic and exhibit the greatest latency.

Internal Business Dimension Instrument the full request chain to capture timestamps between key stages, turning raw data into quantifiable targets.

Reverse‑Order Specialized Problem Solving

Teams form cross‑functional squads to attack specific problems, iterating quickly and documenting reusable solutions, frameworks, and tools. Early projects required 30 people over three releases; later projects (search, core‑link) needed far fewer resources thanks to accumulated tooling and experience.

Long‑Term Forward Control

Continuous governance focuses on standards, processes, and automation platforms. The goal is to prevent performance degradation despite frequent hot‑updates, dynamic plugins, and rapid iteration. Fixed baselines and month‑over‑month metrics are used to quantify changes.

Process Integration

The main development flow includes requirement analysis & design, independent business iteration, and integration testing. Performance checks are embedded at each stage: early detection in design, rapid fixes during iteration, and daily dashboards during integration testing. Post‑release monitoring and alerting close the loop.

Platform Support

Gaode leverages the Titan CI platform and ATap automated testing to create an end‑to‑end toolchain covering build, performance testing, issue tracking, and resolution, dramatically improving detection and remediation speed.

Conclusion

The combined tactical approach (special projects + technical沉淀 + long‑term control) and strategic shift from manual problem solving to architecture‑ and tool‑driven automation have significantly lowered optimization cost and improved efficiency. As tooling matures, further automation may eventually resolve many performance issues without human intervention.