假设您已经在Netflix上看到了诺丁山。然后，流媒体服务建议您接下来观看各种rom-com电影。或者，也许您只是在Kindle上阅读Alex Michaelides的沉默病人。现在，电子阅读器开始向您推广惊悚片。也许您从亚马逊购买了硅胶烘焙垫，肯定的是，下次登录时，亚马逊推荐各种形状和尺寸的烘焙工具。
How did these technologies get so smart that they anticipate what you might need next? The answer is recommender systems (RS). These systems use algorithms to identify related products and audiences based on customer behavior.
A Helpful Overview on eCommerce Recommendation Systems:
Recommendation System, In a Nutshell
推荐系统是信息过滤系统的子类型，该系统可以算出用户如何评价或更喜欢项目。这是一种运行一系列算法，分析数据的工具，有时使用人工智能（AI）提出建议。总体而言，建议系统是向客户建议相关产品的算法。
该系统向每个访问者提供个性化建议，而不是使用机器学习的一般建议。就像经验丰富的营销商一样，他精通销售和交叉销售策略，推荐系统可以帮助增加企业的收入。它根据数据提供建议，包括客户的过去购买，产品评论和用户评分。
These days, a robust recommender system can do double duty as a personal shopper and a shopping assistant. Sophisticated电子商务recommender systems will ensure that the items you want are easily accessible and prominently displayed in your favorite retail stores.
目前推荐系统
We now live in the era of online shopping. But despite the fast growth of eCommerce, brick-and-mortar stores still have the upper hand when it comes to building customer relationships. As a result, online recommendation systems were created to better interact with customers.
亚马逊's recommender system (RS) is widely regarded as one of the best on the market. Enterprise eCommerce solutions likePacvue包括推荐引擎。尽管建议系统的概念似乎是最新的，但事实并非如此。
自计算开始以来，使用计算机选择最佳项目的概念就已经存在。RS概念的首次实现是在1979年以Grundy的形式出现的，Grundy是一个基于计算机的图书馆，向其用户建议书籍。挂毯是第一批广告RS，于1990年代初推出。为了回应人们在互联网上访问和挑选相关数据的努力，挂毯从广泛领域中获取了线索，包括认知科学，模式认识，预测理论，业务管理和marketing technology.
结果，推荐系统作为商机出现，利用可用的信息和专业知识来产生利润。它已演变为任何在线商店的重要组成部分，因为它为用户提供了产品建议，同时还充当虚拟销售人员。这简化了买家的旅程，并优化了他们的整体购物体验。几乎所有流媒体服务，社交网络，在线商店和应用商店都已经使用某种建议算法。
Recommender Systems for eCommerce
您可能已经从亚马逊or another eCommerce shop at some point in your life. And you may have seen a section on the site that says,"Customers Who Viewed This Item Also Viewed..."然后继续招募与原始物品有关的产品。
Source:vwo.com
该部分（在亚马逊上）是推荐系统的一个很好的例子。亚马逊的RS旨在吸引您的业务前景。它还通过建议与您在线商店上最近购买的商品相匹配的产品来提高商店的平均订单价值。例如，如果用户最近通过亚马逊从您的商店购买了枕套，则该网站的RS会建议用户向用户购买毛毯或被子。
The use of recommender systems, sometimes known as product recommendation engines, enables you to better cater to your customer's individual needs by highlighting related products and services. The algorithm will analyze user activity and provide suggestions based on what it thinks the user would like. Users are expected to take note of these suggestions and purchase other items from the set.
The suggested items are recommended based on several factors, such as:
- 商店的畅销书
- 特定类别中最高的项目
- 消费者人口统计
- Purchasing History
- 点击活动
然后，将这些见解利用以预测客户未来的购买模式。使用电子商务推荐系统，如果正确完成，可以提高访客的平均支出并提高品牌知名度。
The “Magic” Behind Recommender Systems
魔术在算法中。有些但不是全部推荐人使用机器学习来确定客户的需求并建议最好的产品。要记住的最重要的事情是RS机器学习取决于一组数据和信息。系统的见解越多，建议就越准确。
The two types of user data supplied to the RS are:
- Implicit: Users generate these pieces of data spontaneously and are related to their navigation behaviors, such as clicks and searches. If a client often buys clothes from your online shop, your recommendation engine will calculate and hint that your buyer is keen on fashion. As a result, the system will mark the buyer's profile with fashion-related tags.
- 显式: Users provide these pieces of information voluntarily provide in response to a question or request for details. They are the feedback left by clients in the form of star reviews and written comments. This form of data is more accessible and transparent.
Information and insights are drawn from a variety of sources. These include the items users browsed, the products they added to their购物车，但后来丢弃，他们的购物历史和搜索变量。这些建议的自定义水平也受平台的意图，收到的数据量以及系统的构建方式的影响。
自定义用户的体验时，系统会使用有关用户过去的操作和首选项（页面访问，用户评分和搜索查询）的信息来提出与用户兴趣更相关的建议。
The acquired data involves three factors:
- 推荐的产品或服务
- 的用户将受到矩形ommendations
- Past Platform Users
最重要的是，建议系统收集数据，作为推论，通过显示和提出选择来简化决策过程。这些选择不仅限于项目。他们可以是一种服务，内容的形式，甚至是一个人或品牌，例如社交媒体上的朋友建议。
The Advantages of eCommerce Recommendation Systems
亚马逊花了多年的时间完善了卢比，这是有充分理由的。如下所示，建议系统提供了各种优势。
Increased revenue.
Recommendation systems boost sales. According to a报告，亚马逊整体购买的35％是由卢比驱动的。已经证明，提供正确的建议选择可以改善销售收入，并使客户购物更好。
将访客转换为买家。
Webstore visitors often explore without purchasing. It's a common customer behavior; they're virtual window-shopping. Through relevant suggestions, eCommerce RS can help visitors find the items they want to buy.
促进交叉销售。
Cross-selling is enabled by recommender systems, which expose buyers to complementary items. If the suggestions are appropriate and helpful, the average transaction size will increase. Your online store can offer more items based on what's inside the buyer's cart at checkout.
Build brand trust.
电子商务recommender systems create value-added and close brand-buyer engagement. Online retailers put effort into understanding their customers, putting this information to use via recommender systems, and developing individualized shopping experiences for each customer. In response, the loyalty of a buyer grows the more they use the recommendation system.
Highlight long-tail items.
长尾商品是难以找到的物品或利基市场。它们非常具体和独特，只有少数人积极寻找它们。RS帮助他们找到不在当地附近的项目，并且他们无法访问其他地方。该系统还可以帮助企业更有效地推广这些项目。
在电子商务推荐系统中过滤类型
Different recommender systems use different data processing techniques. Some RS use data, information retrieval, and pattern matching. Other algorithms integrate these methodologies with AI and machine learning. There are two main filtering types used by the system, regardless of the combination of methods used. They are content-based filtering and collaborative filtering.
基于内容的过滤
这种类型的过滤与用户以及内容或项目之间的功能和相似之处有关。根据用户已经搜索的项目的内容（属性/标签），将显示相关产品。在这种过滤技术中，项目用某些关键字标记。通过内容，系统试图通过浏览其数据库进行潜在匹配来确定用户想要的内容。
当提出最近搜索或购买的物品时，就会发生这种情况。例如，如果客户以前从您的在线商店购买了一部浪漫小说，那么您的RS会假设他们将购买另一本具有类似功能的小说。
|
优点 |
缺点 |
|
No need for other users' data since the recommendation is unique to a particular user |
在某种程度上，项目特征提取是手工设计的，需要熟练 |
|
扩展到大量用户被简化 |
扩展用户当前兴趣的能力有限 |
|
Specific options are based on available history |
协作过滤
协作方法强调用户与产品之间的相互依存关系。构成“相似”项目的东西受客户如何解释不同项目之间的相似性的影响。协作过滤的前提是，过去在某个时候喜欢同一项目的两个人可能会像将来的另一个可比的项目一样。该方法的最好示例是当您看到网站时“带来的客户也带来了”部分。
协作过滤有两种类型。基于用户的基础用户的相似性，而基于项目的相似性依赖于项目相似性。
|
优点 |
缺点 |
|
即使数据有效，有效的工作 |
Cold Start Problem or the difficulty in handling added items |
|
Discovery of new-to-me items based on the past purchasing habits of other users |
有限的侧面功能 |
|
No domain knowledge required |
混合过滤
Many RS are exploring hybrid filtering as a new technique. It incorporates both content and collaborative methods. It provides the system with the capability to understand more precise interactions between people and products. This non-linear filtering technique is also less likely to overstate the user's preferences.
Outsourcing Your Webstore’s RS viaPacvue
亚马逊, the largest online retailer in the world, is a natural case study for anyone looking into recommendation systems. You may remember seeing recommendations tailored to your preferences when you shop on Amazon. When you click on an item, you'll get choices for similar or complementary items. When you add an item to your cart, you'll be able to see what other customers have bought right next to it.
亚马逊的RS是一个成功的故事，主要是因为电子商务网站有很多资源和劳动力来磨练该系统。188金宝慱网站创建自己的推荐系统是可行的，但这并不是公园里的散步，尤其是如果您刚开始零售商店或预算有限。
Fortunately, in today's climate, any virtual shop may have its own system, thanks to dedicated third-party services. Professional assistance tremendously helps to efficiently establish a recommender system in your web shop. Some work on unique projects for each retailer, which takes more time and resources. Others adhere to a standard structure that can be easily customized for any store.
When outsourcing a service for your shop's RS, keep the following factors in mind:
Integration
检查推荐系统是否与您的电子商务平台. If not, figure out if the integration is possible, what steps need to be taken, and if the delay was necessary.
执行
时间就是金钱。一些服务提供商提供了一个更具成本效益的选择，但需要花费时间来设置。其他人则使用更复杂的软件可以扩展，但是它们更昂贵。如果您的预算很小，请找到妥协。您的系统越早操作，您的商店就会越早看到好处。
Cost
In relation to the time of implementation, the price of the system varies between cost-effective, which often uses older software, and expensive, which uses sophisticated technology such as AI.
Analysis
Your provider must have a results analysis tool to monitor the solution daily and accurately. This way, you'll know whether your strategy is working or if you need to adjust tactics.
维护
Check how much maintenance and manual configurations the outsourced RS needs. Fully automated systems decrease personnel expenses and work.
Source:pacvue.com
Pacvue是一个服务,可以帮助你的存储实现recommendation system. Brands and vendors can use the Pacvue platform to improve their digital shelf presence, sales, and marketing initiatives across many marketplaces. Pacvue integrates end-to-end retail data with the resources required to carry out specified actions. The Pacvue enterprise suite assists enterprises and merchants in expanding their operations across Amazon, Walmart, Instacart, Criteo, CitrusAd, Kroger, and other marketplaces.
最后的想法
每个在线商店都需要推荐系统。拥有RS可以帮助您的品牌在电子商务世界中为自己取名，而不论公司的规模如何。电子商务推荐系统不仅有助于增加收入，还可以提高客户满意度，促进品牌忠诚度，并在竞争中保持领先地位。
花点时间检查PACVUE等建议的系统服务提供商，以查看哪种技术最适合您的业务。不管是什么，请确保在您的商店中包括RS。
