Understanding the Mid‑Platform (Zhongtai) Architecture: History, Practices, and Pitfalls in Major Tech Companies

The mid‑platform ("Zhongtai") concept originated in 2009 when Alibaba's shared‑business division was created, later inspired by Supercell’s small, autonomous teams during a 2015 visit, leading Alibaba to launch a "big middle platform, small front‑stage" restructuring. Subsequent adopters include Didi (2015), JD (2018), iQiyi (2018), Baidu (2018), ByteDance (2019), Zhihu (2019) and Tencent (2019), each emphasizing rapid development, data integration, and standardized services.

Major enterprises implement the mid‑platform differently: Tencent builds a technology committee to standardize capabilities; Baidu focuses on reusable, customizable services; Xiaomi structures its platform around business, data, and technology; Didi emphasizes five‑fold "‑ization" (service‑, async‑, config‑, plugin‑, data‑); JD creates a shared, standardized data platform; NetEase distinguishes between "fat" and "standard" platforms; Yonyou evolves its IUAP into a 3+2+1 architecture; Zhihu adopts a "big middle, small front" model with shared technical teams.

The article clarifies terminology: the front‑end (HTML/CSS/JS) renders UI; the back‑end (beans, DAOs, services) handles business logic; the front‑stage combines front‑end and back‑end user‑facing systems; the mid‑platform sits between front‑stage and back‑stage, abstracting core data and services to reduce duplication; the back‑stage manages core resources such as finance, product, and logistics.

Mid‑platforms are classified into three types: data platforms that unify, process, and store massive data for downstream consumption; technical platforms that add business attributes to underlying tech stacks; and business platforms that support online‑intensive services, with the appropriate type chosen based on a company’s primary workload.

Common pitfalls include misaligned team structures, unclear business decomposition, overuse of micro‑services, over‑design, front‑stage trial‑and‑error without governance, profit‑sharing disputes, blind copying of other companies' architectures, and poor leadership selection. Successful mid‑platforms require small, skilled teams, clear ownership, and alignment with both front‑stage product needs and back‑stage stability.

During implementation, organizations often face early survival challenges for the mid‑platform team, staff turnover, sequential building of data and technical platforms before business platforms, risk of the platform’s eventual disappearance if not integrated with front‑ and back‑stage, and decisions about whether to adjust structure first or build the platform first. Effective coordination among all layers and careful organizational adjustments are essential for sustainable mid‑platform success.