Best Practices for Data Tracking Governance at JD Daojia: Process, Standards, and Quality Improvements

JD Daojia, a leading instant‑retail e‑commerce platform, emphasizes the growing importance of data tracking (埋点) and the challenges of its usability and accuracy. The article shares practical experiences in governing tracking data and improving its quality.

1. What is Tracking (埋点)?

Tracking data supports optimization and operational strategies. Common scenarios include product exposure, click, add‑to‑cart, checkout, and order submission, forming a data funnel to assess product, merchant, and brand performance.

2. Why Govern Tracking?

Tracking provides the raw data for big‑data processing, analysis, and mining, yielding key product metrics. However, as business grows, historical tracking schemes reveal issues such as missing process planning, inconsistent design specifications, unstandardized development, poor quality assurance, redundant tracking, and difficulty in usage.

3. How to Govern Tracking

3.1 Establish a Unified Tracking Process

JD Daojia operates on five platforms (Android, iOS, RN, H5, WeChat App). A unified process involves business, product, data product, development, testing, data warehouse, and analysis teams, ensuring each step is well defined.

The new process brings several advantages:

Improved design quality – design efficiency increased by 50%.

Reusable and extensible tracking components, reducing development and testing costs.

Consistent tracking specifications, saving about 30% of data analysis effort.

3.2 Establish Comprehensive Tracking Standards

Standardization covers product requirement specifications, data design specifications, and implementation guidelines.

Product Requirement Specification : All requirements must be submitted via the front‑end product on the tracking management platform, including page, version, description, key parameters, and UI mock‑ups.

Data Design Specification : Four main tracking types are defined – precise exposure (ep), browsing, clicking, and API‑driven events. Example of precise exposure configuration:

"epDuration":1, "epPercent":0.5

Resource identifier format:

res_type|unit_index|res_unit|tpl|index|res_name|sub_page

Click event example (Java implementation):

private void addCart(View v) {
    if (entity == null || context == null || (context instanceof Activity && ((Activity) context).isFinishing())) {
        return;
    }
    if (entity.getIconType() == 1) {
        if (isAllCart(params)) {
            // add to global cart
            addCartSpu(v);
        } else {
            // mini cart handling
            if (context != null && params != null) {
                String userAction;
                if (!TextUtils.isEmpty(entity.getUserActionSku())) {
                    userAction = entity.getUserActionSku();
                } else {
                    userAction = entity.getUserAction();
                }
                DataPointUtil.addRefPar(DataPointUtil.transToActivity(context), pageName, "userAction", userAction, "traceId", traceId);
                new SpuSelectDialog(context, params).showDialog();
            }
        }
        if (onClickAddListener != null) {
            onClickAddListener.onClickSpu(v);
        }
    } else {
        // sku add‑to‑cart handling
        if (isMiniCart(params)) {
            addMiniCartSku(v, params);
        } else if (isAllCart(params)) {
            addCartSku(v, (OnSpuAdapterParams) params);
        }
        if (onClickAddListener != null) {
            onClickAddListener.onClickSku(v);
        }
    }
    if (params != null && params.getPointData() != null) {
        pointData.put("storeId", entity.getStoreId());
        pointData.put("skuId", entity.getSkuId());
        pointData.put("spuId", entity.getSpuId());
        pointData.put("traceId", traceId);
        if (pointData.get("userAction") == null) {
            if (!TextUtils.isEmpty(entity.getUserActionSku())) {
                pointData.put("userAction", entity.getUserActionSku());
            } else {
                pointData.put("userAction", entity.getUserAction());
            }
        }
        DataPointUtil.addClick(DataPointUtil.transToActivity(context), pageName, "click_add", pointData);
    }
}

Additional standards include public parameter extraction, reporting timing, componentization, event naming, field consistency, and event linking across pages and APIs.

3.3 Implement Quality Standards

Quality assurance involves pre‑release testing and post‑release gray‑scale acceptance, covering reporting timing, parameters, scenarios, duplication, redundant events, and empty arrays.

3.4 Optimize the Tracking Platform

The platform now provides a “page catalog”, “resource catalog”, and “tracking catalog” for clear communication between data and product teams, and a data governance portal showing page traffic and user metrics.

4. Benefits of Tracking Governance

Tracking accuracy has risen to an average of 96.5%, exceeding 99% in fine‑grained scenarios, delivering significant improvements over previous practices. The governance model has been adopted by other JD Daojia businesses such as LaiLaiTui, JingMing Manager, and B‑side services.

5. Summary and Outlook

As JD Daojia’s user base expands and new projects launch, data tracking complexity grows, demanding robust quality controls, monitoring mechanisms, and continuous improvement to mitigate risks such as inaccurate reporting, rising manual verification costs, and increased storage and server load.