Douyin Group E‑commerce Data Tracking Evolution, Solutions, and Attribution Practices

The article focuses on the e‑commerce scenario, presenting Douyin Group's data‑tracking (埋点) evolution, solution architecture, attribution practices, and the resulting benefits for data engineers.

1. Tracking Evolution – Before 2013, no log collection was performed; Log 1.0 relied on external tools. Log 2.0 (2013‑2016) introduced a unified event format for recommendation, but failed to launch due to high implementation cost and lack of cross‑team cooperation. Log 3.0 (2017‑present) simplifies the model into events and parameters, adds a flexible point‑management platform, and ensures backward compatibility.

2. Scenario Iteration – As the app grew from a recommendation feed to a comprehensive ecosystem, user paths became deeper and more complex, leading to quality issues (missing/incorrect data), unknown attribution scenarios, and high maintenance costs.

3. E‑commerce Pain Points – Problems appear at four layers: user (incorrect parameters affect recommendations and commissions), analysis (difficulty splitting order sources), engineering (high cost of adding new points across many front‑end teams), and data (parameter explosion, transmission and storage overhead).

Solution Framework – The team built a framework consisting of four pillars: point‑management (BTM/BCM), SDK, attribution platform, and analysis product. BTM adopts a Swift‑Package‑Manager‑like hierarchy (business‑page‑block‑slot). BCM standardizes key dimension fields (product_id, shop_id, etc.) to reduce naming inconsistencies.

SDK Capabilities – The SDK replaces long‑chain URL parameter passing with a BTM‑mode that aggregates and forwards information, handles key event reporting, consolidates parameters, and outputs attribution results. It abstracts user paths for both APP (queue model) and PC/WAP (tree model), enabling reliable reconstruction of user journeys.

Attribution Platform – Built to address delayed attribution updates and undocumented strategies. It tags points, configures corresponding policies, and generates data‑processing tasks via UDFs, providing a unified, maintainable attribution configuration.

Analysis Product – Offers low‑cost module analysis (PV, UV, dwell time, loss rate) and path analysis by extracting and simplifying user paths, supporting detailed user‑level inspection and segmentation.

Benefits – Improved tracking quality (unknown data reduced to < 0.X %), fewer online incidents, higher development efficiency, and enhanced attribution capabilities through accurate user‑path reconstruction.