Algorithmic Optimization for Information‑Flow Advertising at Hello Travel
This talk explains how Hello Travel tackles challenges in information‑flow advertising by describing the market landscape, their business background, and detailed algorithmic optimization across plan, creative, and pre‑bid dimensions, including data‑driven modeling, feature engineering, LightGBM and uplift models, and outlines future directions.
The presentation begins with an overview of information‑flow advertising, its evolution since 2006, and its growing market share, highlighting why Hello Travel uses this format for user acquisition.
It then outlines the advertising workflow from both platform and advertiser perspectives, describing the real‑time bidding process, ad request flow, and the role of DSPs in recall and ranking.
Next, the speaker introduces Hello Travel's business background and the four stages of external advertising development: exploration, initial scaling, cost‑reduction, and maturity, noting that the company is currently in the mature, data‑driven stage.
The core of the talk focuses on algorithmic optimization on the advertiser side, divided into three dimensions:
Plan dimension: Modeling the conversion from certification to order completion using LightGBM, addressing data scarcity and transforming the problem into low‑quality traffic identification and order‑completion probability estimation.
Creative dimension: Building a model to predict whether new creatives can generate volume, employing feature crossing, ID embeddings via word2vec, and handling sparse numerical features to improve budget efficiency.
Pre‑bid prediction: Implementing an uplift model to segment users by conversion intent, enabling precise exposure control and tiered bidding strategies for both new and existing users.
Future directions are discussed, including more precise pre‑exposure screening using exposure data, expanding RTB for user activation, and moving toward fully automated, closed‑loop campaign management with AI‑driven creative generation and cross‑channel optimization.
The session concludes with a Q&A addressing multi‑business competition, creative performance variance across accounts, and the integration of RTB functionality within RTA interfaces.
DataFunTalk
Dedicated to sharing and discussing big data and AI technology applications, aiming to empower a million data scientists. Regularly hosts live tech talks and curates articles on big data, recommendation/search algorithms, advertising algorithms, NLP, intelligent risk control, autonomous driving, and machine learning/deep learning.
How this landed with the community
Was this worth your time?
0 Comments
Thoughtful readers leave field notes, pushback, and hard-won operational detail here.