A Decade of User Growth in the General Entertainment Industry and Its Application to Tencent Video

The article begins with a retrospective of ten years of user growth (UG) in the Chinese general entertainment industry, dividing the period into three stages: 2014‑2017 (the early UG era with 4G rollout and basic traffic acquisition), 2018‑2020 (rapid user base expansion, emergence of short‑video and reading apps, and the introduction of growth‑hacking concepts), and 2021‑present (focus on industrialized, rationalized growth as population dividends wane).

It then discusses how to set reasonable growth targets for Tencent Video, emphasizing a dual‑channel IP strategy that combines platform‑centric user segmentation (based on AARRR metrics) with content‑centric stages (awareness, interest, playback, deep engagement) to align traffic and content objectives.

The resource integration section explains the complexity of traffic attribution, describing first‑party data collection (internal channels, app stores, pre‑installs) and second‑party evidence (last‑click from information‑flow ads, SEM, paid store placements), as well as a hierarchical attribution logic that resolves conflicts between pre‑install and click signals using time‑window parameters.

In the growth‑plan segment, the article details revenue decomposition into membership, advertising, and game‑related income, introducing the DRU (Daily Revenue Users) concept to incorporate paying users without activity into LTV calculations, and describing how non‑standard ads and game‑co‑operation revenues are allocated based on IP exposure depth.

The LTV estimation part outlines four industry‑standard modeling approaches—two‑stage tree models, probability‑distribution models, consumption‑sequence models, and long‑tail segmentation—explaining why Tencent Video adopts a probability‑distribution model for membership revenue and a hybrid classification‑then‑regression approach for advertising revenue.

Effectiveness measurement relies on Gini coefficients and quantile means, while the experimental platform section describes the SOP for evaluating whether a request qualifies for A/B testing, the proportion of experiments that can be run, and the handling of low‑exposure, non‑experimentable, or high‑risk scenarios.

Case studies illustrate the use of causal inference when experiments are infeasible: a pre‑release member‑first viewing experiment that showed significant increases in watch time, and a welfare‑center impact assessment that used propensity‑score matching to demonstrate notable boosts in platform value and user spending.

The presentation concludes with a summary of the four‑step growth framework—goal setting, resource integration, plan formulation, and outcome measurement—and thanks the audience.