Improving Information Exposure Measurement: Visible Ad Metrics and Data Processing Practices at 58 Platform

Background – Traditional information exposure measurement based on retrieval counts suffers from significant bias because it does not reflect whether users actually perceive the information. The current method records server‑side returned lists and client‑side received volume, leading to mis‑valuation of information value and imbalanced training samples.

Ad Visibility – The visibility metric borrows from the Mobile Marketing Association (MMA) standards. An ad is considered a visible exposure when its creative appears in the browser’s viewport and meets pixel‑area and continuous exposure time thresholds. Only visible exposures are counted as opportunities to be seen.

Supplementary Requirements – Measurement must use client‑side logs, filter non‑human or invalid traffic, prevent caching, distinguish intentional page refreshes, and detect focus loss or page occlusion.

Visible Exposure Practice at 58 – The platform defines core requirements: (1) pixel area > 0 px in the visible region; (2) exposure time > 0 s. Supplementary rules enforce client‑side measurement, single counting per retrieval, and no re‑measurement for non‑refresh interactions.

Data Processing – Each information retrieval is assigned a unique TrackId (TID), preferably a UUID, to ensure strong consistency between retrieval and visible exposure data. TID generation and tracking are illustrated in Figure 2.

ETL pipelines unify exposure data from various business lines into a fact table, then join with retrieval facts via TID to produce a consolidated view.

Data Quality – Quality is ensured by monitoring core field fill‑rate, exposure‑to‑retrieval ratios per placement, and the proportion of downstream user actions (clicks, calls, chats) attributable to visible exposures. Alerts trigger when thresholds are breached.

Conclusion and Outlook – Incorporating visible exposure data improves CTR by 7.62 % and CVR by 4.62 %, enhancing user experience and platform stickiness. Future work will focus on expanding measurement coverage, improving data timeliness (moving from T+1 offline to real‑time), and further refining data integrity.

References – MMA China Mobile Internet Advertising Visibility Verification Standard V.1.1.

Author Bio – Lu Yazhou, Senior Big Data Engineer in the Commercial Product Technology Department, responsible for building and developing the site‑wide behavior data warehouse.