This content was originally posted in Spanish here https://blogs.solidq.com/es/poder-del-dato/que-es-el-clustering-segmenta-a-tus-clientes-con-machine-learning/

此内容最初以西班牙语发布在此处https://blogs.solidq.com/es/poder-del-dato/que-es-el-clustering-segmenta-a-tus-clientes-con-machine-learning/

I just introduce the RFM technique to classify your customers in my last article, and I will add next step introducing one machine learning technique such as clustering.

我在上一篇文章中只是介绍了RFM技术以对您的客户进行分类，并且我将添加下一步介绍一种机器学习技术(例如集群)。

How is your customer database composed? Do you collect more information, other than transactional information? Do you use any targeting in your campaigns? Customer segmentation allows you to define a strategy and is a starting point for analysis, lever detection and definition of custom actions for your customers.

您的客户数据库如何组成？ 除了交易信息之外，您是否还收集更多信息？ 您是否在广告系列中使用任何定位？ 客户细分可让您定义策略，并且是分析，杠杆检测和为客户定义自定义操作的起点。

Before proceeding, do not confuse customer segmentation (also known as customer segmentation by related elements) with market segmentation (it provides information on how that market is composed, differentiating groups with similar characteristics and needs). With customer segmentation we identify common features within our customer database. This practice can be common in communication strategies, done manually, based on the observation of these groups and the knowledge of the business.

在继续之前，请勿将客户细分(也称为按相关元素进行的客户细分)与市场细分相混淆(它提供有关该市场的组成方式，区分具有相似特征和需求的群体的信息)。 通过客户细分，我们可以确定客户数据库中的常见功能。 根据对这些群体的观察和对业务的了解，这种做法在交流策略中很常见，可以手动完成。

Searching for common patterns and variables can be a task performed by people… but our capacity is limited by the total information to be processed, whether it is number of customers or number of features to be analyzed. Fortunately the processing capacity of machines is virtually infinite thanks to cloud computing 🙂 What if we combine that knowledge and expertise in the business area with the full potential of Machine Learning?

搜索通用模式和变量可能是人们的任务……但是我们的能力受到要处理的全部信息的限制，无论是客户数量还是要分析的功能数量。 幸运的是，由于有了云计算，机器的处理能力实际上是无限的。if如果我们将业务领域的知识和专长与机器学习的全部潜能相结合，该怎么办？

Whether you’re already working with targeting or want to start segmenting your customer database, read on and find out how clustering can validate those hypotheses or help you discover other types of groups or patterns using the data.

无论您是在进行定位还是要对客户数据库进行细分，请继续阅读并了解聚类如何验证这些假设或帮助您使用数据发现其他类型的组或模式。

什么是集群？ 什么是客户细分？
聚类定义 (What is clustering? What is customer segmentation?
Clustering definition)

Clustering means a set of descriptive (non-explanatory) techniques that aims to form groups from a set of elements, which have different characteristics or variables to allow such grouping.

聚类是指一组描述性(非解释性)技术，旨在从一组元素中形成组，这些元素具有不同的特征或变量以允许进行此类分组。

These groups, focusing on strict classification, must be mutually exclusive; that is, each element must belong only to one group and the groups must be composed of elements as similar as possible and, at the same time, as different as possible between groups. In line with a “traditional” definition, it seeks to achieve homogeneous groups internally and heterogeneous between them (in practice, there are not-so-different boundary cases that can present potential business opportunities).

这些专注于严格分类的组必须互斥； 也就是说，每个元素必须仅属于一个组，并且这些组必须由尽可能相似且同时在组之间尽可能不同的元素组成。 按照“传统”的定义，它试图在内部实现同质的组，并且在它们之间实现异质的组(实际上，没有那么不同的边界案例可以提供潜在的商机)。

This is a statistical process that is usually solved with Machine Learning techniques due to the large amount of data to be processed.

这是一个统计过程，由于要处理大量数据，通常使用机器学习技术来解决。

For its development, we must take some aspects into consideration:

为了使其发展，我们必须考虑一些方面：

• The quality of the data. Sometimes it is essential to standardize such information.
  数据的质量。 有时，标准化此类信息至关重要。
• The total groups to be calculated. It is essential to collaborate with those departments that know the business problem that is intended to be solved to define whether there should be a total of groups to be calculated.
  要计算的总组数。 与那些知道将要解决的业务问题的部门进行协作以定义是否应该计算总计的组是至关重要的。
• The distances between the groups (how different they are) and the size of the cluster, if applicable.
  组之间的距离(它们之间有多不同)和群集的大小(如果适用)。

• The hierarchy when defining those groups. When delimiting classification in a single cluster, making the decision whether it belongs to an “before time” group can condition the classification and be that element in one group less abound than another.

  定义这些组时的层次结构。 在单个群集中划分分类时，决定是否属于“之前”组可以确定分类的条件，并且使一个组中的元素比另一组少。

  Strategic customer segmentation

  战略客户细分

After clustering definition, this concept, a priori, “sounds easy” 😊. In short, it’s about grouping our customer database into different segments where each of them is characterized by certain features or properties that help describe how they are or how they behave. How many variables is it possible to analyze? All of us we have! (*)

在对定义进行聚类后，这个概念(先验)“听起来很容易”😊。 简而言之，这是关于将客户数据库分组为不同的细分市场，其中每个细分市场都具有有助于描述其状态或行为方式的某些功能或特性。 可以分析多少个变量？ 我们所有人都有！ (*)

–> (*) Eye, not the gross… 😉 It is always advisable to debug those that are not relevant and identify those in which there may be interdependencies (we would be giving more weight to these variables than they should be).

–>(*)眼睛，而不是总的眼光……always总是建议调试那些不相关的变量，并找出可能存在相互依赖关系的变量(我们将比这些变量赋予更多的权重)。

Looking for a definition similar to the one described above for clustering:

寻找与上述用于聚类的定义类似的定义：

Strategic customer segmentation means a set of descriptive techniques that aims to form groups from all clients in an organization, which have different characteristics or variables to allow such grouping.

战略性客户细分是指一组描述性技术，旨在从组织中的所有客户组成组，这些组具有不同的特征或变量以允许进行此类分组。

Again, we seek to obtain a certain total of groups, where customers are very similar to each other and, at the same time, sufficiently different from the other segments found.

同样，我们寻求获得一定数量的组，这些组中的客户彼此非常相似，同时又与找到的其他细分市场有足够的差异。

The methodology to be followed will always be linked to the business objective that is intended to be solved. In fact, we could say that customer segmentation succeeds when used by business leaders by providing them with a roadmap or conceptual map on how their customer database is composed, allowing them to make strategic decisions about their actions. Among other decisions, those related to priority and what resources are earmarked for, such as:

遵循的方法将始终与要解决的业务目标相关联。 实际上，我们可以说，当业务主管使用客户细分时，可以为他们提供关于其客户数据库组成方式的路线图或概念图，从而使他们能够制定有关其行动的战略决策，从而成功实现了客户细分。 除其他决定外，与优先级和专用资源有关的决定，例如：

• Cross-selling and up-selling. By identifying purchasing patterns, we will be able to work with product recommendation engines. It will also help maximize the customer’s share, improving their average ticket if we recommend the right products at the end of the purchase process.
  交叉销售和向上销售。 通过确定购买模式，我们将能够使用产品推荐引擎。 如果我们在购买过程结束时推荐合适的产品，这也将有助于最大化客户的份额，提高他们的平均机票。
• Prioritization of leads or acquisition of clones (new customers). Through predictive demand models or information on what our most profitable customers are like, we will be able to invest in target-specific campaigns, or prioritize the attention of those leads, further monetizing the acquisition of new customers.
  优先考虑潜在客户或获取克隆(新客户)。 通过预测性需求模型或有关我们最赚钱的客户情况的信息，我们将能够投资于针对特定目标的广告系列，或优先考虑这些潜在客户，从而进一步利用新客户获利。
• Retention and reduction of abandonment rate. It’s so important to know how to capture those most profitable customers, such as loyalty and retaining them. Working on the satisfaction of our most profitable customers or predicting which groups are most susceptible to abandonment to run specific retention campaigns will be key aspects to reducing our churn rate. –> What is the customer abandonment rate? You might be interested in this article. Look! )
  保留率和减少率。 知道如何吸引那些最赚钱的客户(例如忠诚度并留住他们)非常重要。 努力使最赚钱的客户满意，或预测哪些群体最容易被遗弃以开展特定的保留活动，这将是降低客户流失率的关键方面。 –>客户的放弃率是多少？ 您可能对本文感兴趣。 看！ )

Here are just a few examples of resource optimization. Segmentation or clustering of customers can also help to make strategic decisions related to the optimization of the assortment (identifying the best suppliers or references), the definition of profitability goals for each group of customers, the identification of which channels (or which group within each channel) are most profitable to us or what type of products is most suitable for that channel… and countless information that makes sense when business deciders intervene and are determined to stop deciding on the basis of intuition.

这只是资源优化的一些示例。 客户的细分或集群还可以帮助做出与商品种类优化(确定最佳供应商或推荐人)，每组客户的获利目标的定义，确定哪些渠道(或每组中的哪个组)有关的战略决策。渠道)对我们来说最有利可图，或者哪种类型的产品最适合该渠道……以及无数的信息，这些信息在业务决策者进行干预并决定根据直觉决定停止决策时才有意义。

Need to segment your customers?Already said Philip of Macedonia and then Louis XI: “Divide and conquer”… and although precisely in this article the quote has no war connotations, if you have come this far I hope to have convinced you how important it is to understand how your customer database is composed, whatever your sector or the type of company and that the possibilities to take advantage of that information are many (and of great value for marketing and sales departments) 😊

需要细分您的客户？ 马其顿的菲利普，然后是路易十一世已经说过：“分而治之”……尽管在本文中，报价没有战争意义，但如果您走了这么远，我希望可以说服您了解客户的重要性不论您所在的部门或公司的类型如何，数据库都是组成的，利用这些信息的可能性很多(对于市场和销售部门而言具有巨大的价值)😊

Take the step and make communication between strategic and technical departments take place… and if you don’t have any data yet, put your batteries on! don’t leave it anymore and start collecting information about how your customers are key if you want to benefit from the options presented when working with strategic targeting.

采取步骤，使战略和技术部门之间进行沟通……如果您还没有任何数据，请戴上电池！ 如果您想从使用战略定位时所提供的选项中受益，请不要再犹豫了，开始收集有关客户如何成为关键客户的信息。