当前位置：网站首页>2020-9-14 introduction to advertising system

2020-9-14 introduction to advertising system

2022-06-29 10:19:00 【qqq2018】

The term

RTB： Real time bidding real time biding
Users visit the website - Website according to cookie Track who this user is - Send this user ID to ADX（ Advertising trading platform ）-ADX To multiple DSP（ Demand side platform ） Ask about the price of this ad displayed to this user - The highest price wins , The corresponding advertisement is displayed on the user interface
Advertisers' websites and DSP Also have cookie Map to know which user this is .

UV： Independent visitor Unique Visitor
PV： Page views pageview
GMV： turnover = Number of visitors × Transaction conversion rate of the whole store × Customer unit price

SPU Is commonly known as “ paragraph ”
SKU It's the product “ Pieces of ”

CTR（Click-Through-Rate） Click through rate = Click on / Exhibition

Business logic

technological process

With ad-server（ Advertising machine ） Centered , Other modules are related to ad-server communication .

Flow through gateway Parse user information into ad-server Enter after the required format ad-server.
ad-server Send user information to user-server Find the user's historical behavior （ Search for 、 Browse 、 Shopping records ）.
ad-server According to the user's historical behavior to sku-server Find the product information corresponding to the historical behavior in （ Name of commodity 、 brand 、spu、 Level III directory, etc ）.
ad-server Enter the above information into re-server（recommend-server） and ad-retrieval（ Ad search ）, Return the result to gateway Used for front-end display .

re-server Returns the... Generated according to the recommendation logic sku list （ How to sort the list depends on the sorting model ： First feature-server Extracting features , And then use predictor Predict which are more likely to be bought ）
ad-retrieval function ： When online, search the advertisement candidates with composite conditions from the advertisement index according to the user tag and page tag .

Focus on modules

gateway

All the traffic at the front end reaches gateway.
be responsible for ：
Front to back ：
shunt （abtest、 Divert some to the small flow experimental platform ）
Keyword penetration （ front end ->gateway-> Back end , Used to track user behavior , Billing, etc ）
Back to front ： Serialize the advertisement into the format required by the media

mixer

What we do is recommend advertisements （ Display advertising ）

Predictor

Real time indicators for predicting online advertising （ Not predicting what customers like ）
CTR:Click Through Rate Click through rate
pCTR：predict CTR
eCPM： A thousand times show expectation （expected） income

Basic services

Hadoop

from Apache Distributed system infrastructure developed by the foundation . Users can do this without knowing the underlying details of the distribution , Develop distributed programs . Make full use of the power of cluster for high-speed operation and storage .
Hadoop Implemented a distributed file system , One of the components is HDFS.HDFS It has the characteristics of high fault tolerance , And designed to be deployed on low-cost hardware ; And it provides high throughput to access application data , For applications with large datasets .

flume

Highly available , Highly reliable , Distributed massive log collection 、 Converged and transported systems ,Flume Support for customizing various data senders in the logging system , To collect data ; meanwhile ,Flume Provides simple processing of data , And write to the various data recipients （ Customizable ） The ability of .

KAFKA

High throughput distributed publish subscribe message system , Persist messages to disk .
It can be used for mass consumption and real-time applications