New retail industry search best practices

Introduction to This article uses new retail customer cases to introduce you to the industry characteristics of retail e-commerce and fresh food e-commerce online business search, and how to build intelligent search services through the open search e-commerce enhanced solution to quickly implement various items The improvement of indicators has brought more new opportunities to the business.

Customer background

A well-known supermarket retail brand has more than 10,000 stores in more than 10 countries around the world. In addition to offline stores, it also provides online shopping malls, which can enjoy timely delivery services and create a multi-faceted shopping experience for consumers.

In the context of the rapid development of the domestic new retail market, online business is particularly important if you want to quickly deploy and enhance brand influence. Among them, the search function that can directly affect user experience and business conversion is facing challenges. has quickly built a high-quality search service in just 3 months since the survey to access the open search, and through the optimization of the search capabilities of the enhanced version of the e-commerce industry, the improvement of various indicators has been achieved, which has brought more to the business New opportunities.

Search business background

Customer self-built search pain points

1. R & D angle:

• Lack of industry lexicon accumulation, word segmentation and query analysis without industry feature adaptation, lack of a large number of industry sample data, and difficult self-study;
• A mature search engine involves offline modules, online modules, query understanding services, algorithm platforms and other system components. It requires a lot of development, algorithm tuning, and continuous complex operation and maintenance work, and the cost of self-built is high;
• The development cycle is long and it is difficult to respond quickly to business changes;

1. User angle:

• Can't find it, don't want to buy it, users are lost;
• Poor search relevance, low conversion rate, and poor experience;
• I found it, ranked behind, it is not convenient to purchase, and the experience is poor;

1. operational point of view:

• Search effect directly affects business conversion;
• The thesaurus is incomplete/outdated, the sorting effect does not meet the business needs, and the operation intervention cost is high;
• Lack of search-related real-time operation reports to support data analysis and business decision-making;

Business characteristics and needs

1. Commodity brand category SKU is various, update fast;

• Searching for "milk" has recalled dozens of brands, each of which contains different categories of dairy products;
• Regularly update new products on the shelves and remove old products from the shelves;

1. same category have diverse semantics and many classifications;

• Searching for "beef", the intention may include: sirloin, steak, beef tendon, sauced beef, beef rice, and other sub-categories;
• "Cabbage" = cabbage, green cabbage, kohlrabi;

1. same product belongs to multiple categories;

• "Yogurt" belongs to the category of dairy products as well as the category of casual snacks
• "Tomato" belongs to the category of vegetables as well as the category of fruits
• "Beef" belongs to the category of beef and mutton, but also belongs to the category of cooked food, meatballs, etc.

1. Commodity name contains Chinese+English/number combination;

• "Swisse Liver Protection Tablets", "Laundry Liquid 2L"

1. search filter conditions are complex;

• The chain store search needs to return the most suitable products according to the conditions such as whether there is inventory, whether there is promotion, whether new products, sales, whether on shelves, etc.;

1. needs to build a personalized search guidance function;

• Algorithm models such as shading, hot search, and drop-down prompts

1. Recall sorting rules flexible intervention adjustment;

• Better integration of self-operated products and products of other brands, and the results of designated product recalls are on top;

1. query and analysis functions such as word segmentation, synonyms, spelling error correction, stop words, etc. flexibly intervene;

• Intervene the product entry "milk powder s3=milk powder 3 segments" according to the characteristics of the product,

Open search for enhanced solutions for the e-commerce industry

Introduction to Open Search

OpenSearch (OpenSearch) is a one-stop intelligent search business development platform built by a large-scale distributed search engine independently developed by Alibaba Cloud. No development is required, and high-quality search services can be obtained with one-click access. The built-in Alibaba technology has accumulated for many years. Core search engine, industry-leading search capabilities and algorithm capabilities, and fully open to support internal calls to customers' own algorithm models to meet the business needs of various industries and scenarios, and achieve mutual achievement and common growth with customers;

Enhanced search architecture for the e-commerce industry

The enhanced version of the open search e-commerce industry is the industry's first solution. With years of business experience and tens of thousands of customer services, through research and analysis of different scenarios and industry search characteristics, it is implemented through industry template productization.

solution

1. built-in e-commerce industry analyzer, one-key configuration;

• Query analysis functions: e-commerce error correction, e-commerce word weight, e-commerce stop words, e-commerce synonyms, e-commerce word segmentation, e-commerce entity recognition;
• Sorting strategy: e-commerce sorting expression;
• Search guide service: drop down prompt

1. Industry-leading search technology-the application of product search in multiple recalls

1. E-commerce exclusive sorting expression, access to get high-quality sorting effect

• Two rounds of sorting are supported. Basic sorting (coarse sorting) is audition selection. You can quickly find high-quality documents from the search results, take out TOP N results and then perform fine calculations according to business sorting (fine sorting), and finally return the best The result is to the user.

。

• The category prediction model improves the sorting effect. The higher the correlation between the category and the query, the higher the ranking score the item gets, so the item will be ranked higher.

1. built-in higher quality algorithm model, saving training costs;

The system directly built-in Tao system search algorithm capabilities, supports multiple ways of uploading behavioral data, and automatically trains the algorithm model daily: hot searches, shading, drop-down prompts, etc., to achieve personalized search services,

1. query analysis intervention dictionary;

Example: User search Query: Milk powder s3900g

• • Participle: milk powder/s3/900/g
  • Synonym: "milk powder s3=milk powder 3 segments"
  • Stop word: "gram"
  • Word weight: milk powder-medium, s3-high, 900-high, grams-low

1. category predictive intervention dictionary;

Example: The user searches for Query "Bread", and the returned results think that "Star Ou Bao" is strongly correlated, and "Toast" is weakly correlated;

• • Solution: Intervene the category relevance of Query "bread" in the category prediction dictionary

1. 1613986cc40093 satisfies the participation of non-technical students on the operation side to

• Support the creation of custom permissions for sub-accounts, meets the permission requirements of different positions ;
• The console has a complete data management capability , which can directly view reports such as business operation reports, and make corresponding operation analysis and decision-making based on the search business index data;

Voice of customers

From mid-July to the end of the month, all stores are connected to open search, which has a significant effect on business conversion in just half a month;

1. solves the technical difficulties of self-built search

• Taobao and Tmall thesaurus is the same, the unique e-commerce industry attributes are more suitable for business needs, and the thesaurus is continuously updated;
• Open search can adjust the sorting rules at any time, and the sorting dimensions are richer. It can also affect the sorting in real time according to the inventory changes. It easily solves the problems of the program control sorting logic in the previous self-built, and the inventory changes require index reconstruction to affect the sorting.
• Satisfy the personalized ranking needs of the business, such as: the search results specify the top of the product, and the machine learning algorithm automatically predicts the relevance of keywords and categories;
• Multi-channel recall is realized through open search Chinese and English index + Zhongtai original index;
• In the past, the search solution was not very likely to be upgraded, and now it can enrich its business and grow better based on the continuous iteration of the open search product capabilities;

1. Reduce IT operation and maintenance costs

• Cloud fully managed service, visual configuration, distribution of node responsibilities in the cluster, cluster load, number of index shards and number of replicas are all mature tuning mechanisms. Computing power can be flexibly scaled as needed, ensuring service performance without additional investment Machines and operation and maintenance personnel do not need to do centralized optimization and index optimization by themselves, and the overall cost-effectiveness is higher

1. Improve operational efficiency

• Rich e-commerce thesaurus reduces the workload of manual maintenance;
• The operation side can directly participate in the intervention and tuning of the console, which greatly accelerates the response of the business;
• Rich data analysis dimensions assist operators in continuous optimization and enhance user experience and conversion;

1. Various business indicators improved

• The overall conversion rate of additional purchases increased by 10%;
• The search result rate has been reduced from a recent peak of 29% to 7.5%;

---

If you need product guidance, you can fill out the questionnaire to get expert guidance \>> https://survey.aliyun.com/apps/zhiliao/lKD\_J8cRj

If you want to communicate with more developers, understand the cutting-edge search and recommendation technology , you can scan the code to join the community

Copyright Notice: content of this article is contributed spontaneously by Alibaba Cloud real-name registered users. The copyright belongs to the original author. The Alibaba Cloud Developer Community does not own its copyright and does not assume corresponding legal responsibilities. For specific rules, please refer to the "Alibaba Cloud Developer Community User Service Agreement" and the "Alibaba Cloud Developer Community Intellectual Property Protection Guidelines". If you find suspected plagiarism in this community, fill in the infringement complaint form to report it. Once verified, the community will immediately delete the suspected infringing content.