案例库 · 软件与 IT · 产品决策 · 2010-2014
这条还没译成中文，下面是英文原文。
Twitter's Who-To-Follow engine created 1 in 8 new connections
An algorithmic recommendation system decided whom to suggest, drove growth, and became the foundation of Twitter's ad business.
那一手
Twitter's growth depended on new users finding accounts worth following, but an empty timeline gives a new user no reason to stay. The company built Who-To-Follow (WTF), an algorithmic product that recommends accounts, solving algorithmic, operational and experimental challenges along the way.
WTF was treated as a data product: recommendations were computed at scale offline and served in real time, then measured with experiments. More than one-eighth of all new connections on Twitter are a direct result of the system, and it materially improved engagement quality.
The same system became a foundation for promoted products, Twitter's ads, which produced a substantial majority of the company's revenue. At publication, Twitter was publicly traded with a market capitalisation above $30 billion, close to $1 billion in projected annual revenue, and over 240 million active users.
为什么管用
- It attacked the root cause of new-user churn: an empty follow graph.
- Every accepted suggestion adds a durable, compounding connection.
- The underlying machinery doubled as ad-targeting infrastructure.
- Experimentation let the team tune recommendations by measured impact.
可以搬走什么
If your product's value is the network, invest in the machinery that adds edges. The same recommendation system can later power monetisation, so the growth asset becomes the revenue asset.
后来呢
WTF remained central to Twitter's onboarding and ad products; the 2015 Interfaces paper documenting it is a canonical applied-OR case study. Twitter later renamed itself X, but the recommendation-plus-ads pattern persists across the industry.
资料来源
- The Who-To-Follow System at Twitter: Strategy, Algorithms, and Revenue Impact
- The Who-To-Follow System at Twitter: Strategy, Algorithms, and Revenue Impact
发现哪里写错了？告诉我们。