How knowledge atlas, graph data platform and graph technology help the rapid development of retail industry

Graph query  |  Data warehouse  |  Figure data platform

Graph algorithm  |  Data expansion  |  Graph data science

The official account Series in the last issue showed you what is graph database and the ten fields of graph database application . Let's share with you :

One 、 Current data trends and challenges

Two 、 What are the categories of knowledge maps

3、 ... and 、 What are the ways of data storage

Four 、 What are the benefits of adopting a graph data platform

5、 ... and 、 How to use graph technology , Innovative retail development

Current data trends and challenges

Data isolation

At present, some data are isolated from each other . Data storage and applications usually serve a single department . For example, the human resources department uses a platform , The sales department may use another platform .

Data expansion and data lake

In the context of deep learning , Most data has a data lake 、 Data warehouse 、 Relational database , As a recording system 、 Customer data 、 Trading data 、 Product data and order data . This data dispersion will lead to data expansion .

The data lake can store a large number of structured data 、 Semi structured and unstructured data , And the cost is low , So it's very popular .

In terms of cost , Data lake has great attraction , It plays an important role in storing any type of data , Including log files generated by applications and services . The operation of saving data into the data lake is simple and convenient . However , Managing and understanding this data is difficult .

Cloud storage

Cloud computing is disruptive , But there are still management challenges , This is mainly due to more data stored in more systems . Personal cloud data may be stored in  iCloud、Google Drive、Dropbox、Evernote、Gmail  and  Notes  in . 

At this time, a large amount of data needs to be processed , These data are not only in different formats , And part of the data is repeated , And most of the data is not correlated .

Historical data will be forgotten

Historical data provides power for machine learning prediction . With the outbreak of the COVID-19, it caused economic turmoil , Cause historical data to become obsolete . For example, historical data is often used to predict purchasing behavior . However, due to the closed state during the epidemic , Online shopping has become a market leading way of shopping , Buying behavior changed almost overnight . As consumer behavior has undergone earth shaking changes , Therefore, historical data can not accurately predict the purchase behavior .

With limited data , The importance and value of data association are increasingly prominent . Data association and data relationship can be captured by storing data into graph data platform . Of course, historical data is valuable , But storing data and existing relationships between data can improve predictive power , Even in the absence of relevant historical data .

This is because data association and relationship are the most predictive elements in data . All the above factors are driving the enterprise to transform to the data in the association diagram data platform , So as to acquire knowledge .

Knowledge map category

The knowledge map is like knowledge itself , It's all about , Wide range of information . Generally speaking , Knowledge maps fall into two categories ：

Action knowledge map and decision knowledge map .

Action oriented knowledge map

Data management is an important use case of knowledge map . Many well-known enterprises use knowledge maps to create metadata centers , And use this to capture data lineage ： data source 、 Conversion mode and cleaning mode . Knowledge atlas is used to model complex data transmission pipelines , So that consumers and producers of data can be easily identified ,  And integrate new data sources .

With a strong foundation of data sources , Enterprises can take action against the collected data , Accurately understand the data source 、 Data producers and consumers .

In addition to data management , The action knowledge map is also used for personalized recommendation . The actionable knowledge map brings together all relevant data such as customers and products into one  360  In degree view , This will drive a lot of action , For example, identify customers at risk of loss and provide effective suggestions that can persuade customers to stay .

A decision-making knowledge map for data analysis

Knowledge maps form the basis of modern data and analysis . Data captured with knowledge maps , Capture and store all the relationships inherent in the data , There is no need to guess the data correlation . thus , The knowledge map represents a more faithful representation of the data , And enable enterprises to unlock their predictive power .

With the help of decision-making knowledge map , The ultimate goal is to make better decisions , Whether the decision comes from human decision or algorithm decision . These decisions can be supported by .

Graph query

You can answer any questions about the knowledge map in batches .Boston Scientific  Use advanced queries to analyze root causes , And identify the faulty component combination that caused the defect （ This is a reverse recommendation ）.

Graph algorithm

Recognize patterns in data , Such as the shortest path between two points or the most influential customers .

OrbitMI  Use decision-making knowledge map to implement complex container shipping route planning . Through the routing algorithm , They mapped out the sea route in less than a second . Besides , Their knowledge map also supports  SaaS  Analyze the product . The knowledge map will not only produce economic efficiency , And it can also bring efficient and complex line planning , And reduce 6 10000 tons of carbon emissions .

Graph query

and

Graph algorithm

It can also unlock the prediction function of machine learning . AstraZeneca uses graph algorithms and machine learning in its knowledge map to identify patient medical record prototypes and patterns . This study enabled the company to identify strong contacts for early intervention , So as to improve the therapeutic effect of kidney diseases .

What are the data storage methods ？

as everyone knows , The data has  “ potential ”  value , Some companies can extract more value from data than others . To efficiently transform data into information 、 Knowledge and value , There are many drivers to consider , Such as data-driven leadership 、  Knowledge and skills mastered by the company's employees 、 Business processes and organizational culture . But in the process from data to insight , Another important driver is data storage technology . The process of extracting value from data depends on how it is stored 、 Ease of access , And data and other relevant data within the organization 、 The relevance of information and knowledge scenarios .

There are three common types of database management systems ：

Relational database management system

Store data in a table . When the relationship between data is stable and can be captured in the top-level table definition , This type is often used . Relational databases can quickly process large numbers of records , Less storage space is used . But when dealing with dynamic data that the scene may change over time , The flexibility is slightly insufficient .

Non relational database management system

Provide a variety of non tabular alternatives to organizing 、 management 、 Store and retrieve data . Traditionally , Non relational database management systems are optimized for specific data types , It's usually an old system , Difficult to handle updates 、 More dynamic workloads .

Dynamic data management system

Than  RDBMS  More agile and efficient , More attention , As it matures and overcomes potential problems ,2016-2019  Of  CAGR  The growth rate reaches  83.2%. Graph database is a dynamic database management system . With the increasing demand for graph driven analysis and artificial intelligence tools , The demand for graph database and graph database platform is also increasing .

Figure what benefits the data platform brings ？

Reveal hidden structures

Graphs can reveal hidden structures in data , Where the data is unknown , No extensive data storage is required , Nor does it require a lot of prior knowledge about how data is organized .

Increase of efficiency

In some cases , Figure database management system can be better than traditional  RDBMS  faster 、 More effective ,  Because they can quickly analyze patterns and relationships . The graph database also simplifies extraction - Execution cycle ,  You can execute queries faster .

Wide range of functions

Graphs can be applied to various problems and applications . These include ： 

• Discover people 、 A pattern of relationships between places or things  

• Map or draw the structural spatial relationship between the positions of things  

• Capture semantic structure from human language in order to systematically understand content based on pedigree 、 Space based and law enforcement based / Intelligence services “ Known associations ” analysis

• Pharmacological research 、 Epidemiology and utility network analysis

Using artificial intelligence

Graph is likely to play a key role in future artificial intelligence . Because actions and consequences in complex systems often lead to changes in data relationship patterns , So graph databases are helping to drive machine learning and other  AI  Relevant operations realize innovation .

Enterprises understand how to transform data into information in the organization 、 The core approach to knowledge and business value . Identify how map databases and platforms can help enterprises cope with challenges databases are helping organizations transform data into information 、 Knowledge and business value play an important role . Understand the benefits of graphs , Use these advantages to improve the learning ability of the organization .

The blue ocean brain map data all-in-one machine has a simplified and highly available cluster architecture . Soft and hard integration , Highly integrated . Open the box . Better than the current centralized storage architecture X3, Higher than centralized storage architecture X5. Professional operation and maintenance platform , In depth monitoring and management all-in-one system . Distributed storage , high reliability , Full architecture redundancy design , Avoid any single point of failure , And cross node data protection , Better serve all walks of life .

The retail industry is innovating rapidly by drawing technology

In today's society , Retailers face many new and complex challenges . Due to low cost and high sales , Amazon and other online giants quickly distribute products at a lower cost , This has led to successive failures of smaller retailers .

To get a foothold , Retailers must be flexible enough , We have to face huge online competition , We have to deal with another new situation in the retail industry ： Customers are now at the center of the value chain . In order to adapt to these situations , Retailers must control inventory in real time 、 Payment and distribution systems . However , For traditional retailers who are burdened by traditional infrastructure , Real time response is difficult .

In order to redesign the highly correlated value chain from linear to circular , Retailers need to modernize their infrastructure quickly and cheaply . Besides , Internet based online retailers must find a way to deal with scale and complexity , To maintain a competitive advantage .

Let's take a look at how offline retailers make use of

Figure data platform technology

To meet these pressing challenges .

Personalized products and promotional recommendations

Maximize revenue by providing real-time recommendations to online customers , It can improve the customer experience , And increase sales . However , What customers want is a well-designed recommendation , Instead of rigid or blind recommendations . To improve the effect , Recommendations must be based on consumer preferences 、 Shopping records 、 Interests and needs are personalized .

Real time recommendation needs to associate a large number of complex buyer and product data （ As well as the regular associated data ）, To understand customer needs and product trends . This cannot be achieved through relational database technology , because SQL The query will be very complex , And it takes too long to provide recommendations in real time .Hadoop  and  Spark  And other big data processing technologies are also facing the same problems , These techniques work well in areas such as email recommendations , Send recommendations once a day , But not in real time .

After design , Figure the data platform can quickly query customers' purchase records , And immediately capture any new interests displayed in their current online access , Both are key to providing real-time recommendations . Because the relationship is regarded as the first level entity in the graph data platform , Retailers can associate customers' browsing records with purchase records and offline products and brands . thus , The real-time recommendation algorithm can make use of customers' past and present choices to provide personalized recommendations . There is no need to calculate offline in advance , Thus, the delay problem can be well solved .

By the customer  360  Provide a personalized experience of support

Retailers can according to the wishes of customers 、 Interests and needs provide relevant content , To personalize the online customer experience . This will not only increase customer engagement , It also increases revenue and customer loyalty . for example , By providing relevant blog posts next to the product description , Retailers can portray themselves as experts in using specific products . thus , Customers think they get valuable information from reliable sources , So increase access and purchases . 

Retailers can also use path analysis to help improve performance , This includes analyzing the customer behavior that led to the purchase , And use these data to attract customers to a more profitable path . This may require adjusting the content , Or change the target location for future customers to go to after clicking the link .

Retailers can also identify dimensions shared by a group of customers , And cluster them according to these attributes . for example , Customers can be grouped around the attribute of having or not having children , It can also be grouped according to occupation and position , For example, an engineer in his early career and an experienced vice president of marketing . Consumers of different dimensions have different responsibilities and incomes , So buying habits are different . Retailers can use this information to provide personalized content for each customer .

Retailers have a lot of data , It can be used to determine the best path and content to provide services to customers . These data are related to products 、 market 、 social media 、 Master data 、 Data related to digital assets . However , These data are usually stored in information islands , This makes it more difficult to integrate and identify opportunities to provide relevant content to customers . 

To merge all data sources into one personalization engine , Relational databases will not be able to perform complex recommendation calculations in real time . Enterprises can move data to  Hadoop  Or data warehouse , In order to calculate the recommendation for each customer in advance , But these recommendations are usually a little out of date . Besides , If only a small number of customers visit the website on any given day , Calculating recommendations in advance for the entire customer base every day will reduce efficiency .

Figure the data platform does not centralize all customer data into one system , Instead, you leave the data where it is , And add graph analysis . You can provide a department identifier for each customer , Then bind it to the primary customer identifier . The identifier of each department or line of business consists of individual identifiers , To generate a double overlay for each customer . This enables retailers to have a more comprehensive understanding of customer relationships , And quickly navigate back to the original system at any time when the customer interacts with the company .

Use graph technology to optimize e-commerce distribution service route

Amazon has set shipping and distribution standards . Because Amazon  Prime  Members can enjoy free shipping for two days , E-commerce shoppers are reluctant to wait for more than two days to receive online shopping goods . therefore , Retailers must meet or exceed this standard , Otherwise, it may give customers to Amazon .

In order to shorten the delivery time , Retailers must understand the inventory and transportation network of stores and distribution centers . for example , We need to know whether the routing problem will lead to the delayed delivery of products from the distribution center close to the customer , Or whether the product shortage will cause the product not to be delivered on the specified delivery date . Determining the fastest distribution route needs to support a large number of complex route queries , And has fast and consistent performance .

Because the data is highly correlated , The e-commerce distribution service route and map data platform is like “ A heaven-made match ”. This is not just because a large number of... Are required between data points “ Jump point ”, And there can be many different paths and any number of permutations . Even in an order , These permutations may also be optimized , And become the best path for different products at different times of the year . Figure the data platform can take these different factors into account , And support complex route query , To simplify distribution services .

Supply chain visibility

The supply chain is huge and complex . Products usually consist of different raw materials or components , These parts circulate among different suppliers , Each of these components may consist of sub components , The sub components may come from other sub components and other suppliers around the world . Because of this complexity , Retailers often only know their direct suppliers , This may lead to risk and compliance issues .

Retailers need to know the whole supply chain openly and transparently , To detect fraud 、 Pollution 、 High risk locations and unknown product sources . for example , If a particular raw material is damaged in some way , The company must be able to quickly identify each affected product . This requires managing and searching large amounts of data without latency or other performance issues . Transparency is also important for identifying weak links or other single points of failure in the supply chain . for example , Previously a part or ingredient was available from three suppliers , But now it can only be obtained from one supplier , So retailers need to understand the impact this may have on future production .

To achieve visibility throughout the supply chain , Deep connections need to be established . Relational databases are not designed to handle large numbers of recursive queries or associations , So performance is affected . However , Graph data platform is specially designed for searching and analyzing related data . Figure the architecture of the data platform is first built around data relationships , This enables retailers and manufacturers to manage the search of large amounts of data without performance problems , And achieve the supply chain visibility they need .

Improve revenue management with the help of graph technology

For consumers , Shopping around has never been easier than it is today . In a few minutes , Consumers can compare the prices of specific products among a dozen stores , And the whole process can be completed easily . They can even compare prices when shopping in physical stores of different retailers , And buy from competitors . To fight a price war and optimize profitability , Retailers need to provide competitive prices in real time . 

Competitive pricing is based on inventory 、 place 、 season 、 Consumer demand and other factors , These factors are very unstable , It changes quickly . for example , If a hotel plans to price according to the basketball tournament , And there are seven matches in this championship , Then the inventory of the cities hosting the Games will be reduced , And pricing accordingly . But if the tournament ends after five games , Then the inventory of the last two games will increase , And the price should be adjusted properly .

Besides , Each retail outlet may set different prices according to market conditions . The more retailers know about their micro market , And optimize product pricing to match inventory , There are more ways to improve profit margin and sales . However , Relational databases can't keep up with the pace of these data changes , And poor performance makes it impossible to provide real-time price updates across multiple locations . 

Figure data platform can help retailers solve the problem of revenue management , It also provides the processing power and performance required by the real-time pricing engine . The interdependencies between many variables can be represented by graphs , This provides retailers with a way to determine and efficiently calculate prices , Even when this dependency changes rapidly .

System administrator ： The Internet and  IT  operating

retail  IT  Organizations also benefit from the graph data platform . Usually , The company has a complex network , And more and more components are stored in the cloud or multiple clouds and internal data centers . In most traditional configuration management databases  (CMDB)  in , To represent each item  IT  It is difficult to understand assets and how they are related . 

To run multiple virtual machines  (VM)  For example, the physical server of , these  VM  It may host containers that run different processes and are associated with different subnets . under these circumstances , You can use the graph data platform to view the association mode of all components .

System administrators can also use the graph data platform to maintain graphs composed of all different network assets . This diagram can be used to better protect the network and detect vulnerabilities or limit the spread of intrusion risk .

Retailers face many challenges , A large part of this comes from technology savvy online retailers . From providing real-time product recommendations to   State pricing and optimization of distribution routes , Retailers must quickly overcome these challenges , To gain a firm foothold and competitive advantage .

Besides , Retailers must also be more responsive , To keep up with changing consumer and technology trends ahead of competitors . The graph data platform can help retailers understand related data , And compared with traditional technology , Ability to leverage data relationships more deeply , To provide real-time product recommendations 、 Optimize distribution route capability