抽象類細分舉行

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/

此內容最初以西班牙語發布在此處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紹了RFM技術以對您的客戶進行分類，并且我將添加下一步介紹一種機器學習技術(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.

您的客戶數據庫如何組成？ 除了交易信息之外，您是否還收集更多信息？ 您是否在廣告系列中使用任何定位？ 客戶細分可讓您定義策略，并且是分析，杠桿檢測和為客戶定義自定義操作的起點。

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.

在繼續之前，請勿將客戶細分(也稱為按相關元素進行的客戶細分)與市場細分相混淆(它提供有關該市場的組成方式，區分具有相似特征和需求的群體的信息)。 通過客戶細分，我們可以確定客戶數據庫中的常見功能。 根據對這些群體的觀察和對業務的了解，這種做法在交流策略中很常見，可以手動完成。

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?

搜索通用模式和變量可能是人們的任務……但是我們的能力受到要處理的全部信息的限制，無論是客戶數量還是要分析的功能數量。 幸運的是，由于有了云計算，機器的處理能力實際上是無限的。if如果我們將業務領域的知識和專長與機器學習的全部潛能相結合，該怎么辦？

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.

無論您是在進行定位還是要對客戶數據庫進行細分，請繼續閱讀并了解聚類如何驗證這些假設或幫助您使用數據發現其他類型的組或模式。

什么是集群？ 什么是客戶細分？
聚類定義 (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.

聚類是指一組描述性(非解釋性)技術，旨在從一組元素中形成組，這些元素具有不同的特征或變量以允許進行此類分組。

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).

這些專注于嚴格分類的組必須互斥； 也就是說，每個元素必須僅屬于一個組，并且這些組必須由盡可能相似且同時在組之間盡可能不同的元素組成。 按照“傳統”的定義，它試圖在內部實現同質的組，并且在它們之間實現異質的組(實際上，沒有那么不同的邊界案例可以提供潛在的商機)。

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

這是一個統計過程，由于要處理大量數據，通常使用機器學習技術來解決。

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

為了使其發展，我們必須考慮一些方面：

• The quality of the data. Sometimes it is essential to standardize such information.
  數據的質量。 有時，標準化此類信息至關重要。
• 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.
  要計算的總組數。 與那些知道將要解決的業務問題的部門進行協作以定義是否應該計算總計的組是至關重要的。
• The distances between the groups (how different they are) and the size of the cluster, if applicable.
  組之間的距離(它們之間有多不同)和群集的大小(如果適用)。

• 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.

  定義這些組時的層次結構。 在單個群集中劃分分類時，決定是否屬于“之前”組可以確定分類的條件，并且使一個組中的元素比另一組少。

  Strategic customer segmentation

  戰略客戶細分

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! (*)

在對定義進行聚類后，這個概念(先驗)“聽起來很容易”😊。 簡而言之，這是關于將客戶數據庫分組為不同的細分市場，其中每個細分市場都具有有助于描述其狀態或行為方式的某些功能或特性。 可以分析多少個變量？ 我們所有人都有！ (*)

–> (*) 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，而不是總的眼光……always總是建議調試那些不相關的變量，并找出可能存在相互依賴關系的變量(我們將比這些變量賦予更多的權重)。

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

尋找與上述用于聚類的定義類似的定義：

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.

戰略性客戶細分是指一組描述性技術，旨在從組織中的所有客戶組成組，這些組具有不同的特征或變量以允許進行此類分組。

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.

同樣，我們尋求獲得一定數量的組，這些組中的客戶彼此非常相似，同時又與找到的其他細分市場有足夠的差異。

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:

遵循的方法將始終與要解決的業務目標相關聯。 實際上，我們可以說，當業務主管使用客戶細分時，可以為他們提供關于其客戶數據庫組成方式的路線圖或概念圖，從而使他們能夠制定有關其行動的戰略決策，從而成功實現了客戶細分。 除其他決定外，與優先級和專用資源有關的決定，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.
  交叉銷售和向上銷售。 通過確定購買模式，我們將能夠使用產品推薦引擎。 如果我們在購買過程結束時推薦合適的產品，這也將有助于最大化客戶的份額，提高他們的平均機票。
• 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.
  優先考慮潛在客戶或獲取克隆(新客戶)。 通過預測性需求模型或有關我們最賺錢的客戶情況的信息，我們將能夠投資于針對特定目標的廣告系列，或優先考慮這些潛在客戶，從而進一步利用新客戶獲利。
• 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! )
  保留率和減少率。 知道如何吸引那些最賺錢的客戶(例如忠誠度并留住他們)非常重要。 努力使最賺錢的客戶滿意，或預測哪些群體最容易被遺棄以開展特定的保留活動，這將是降低客戶流失率的關鍵方面。 –>客戶的放棄率是多少？ 您可能對本文感興趣。 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.

這只是資源優化的一些示例。 客戶的細分或集群還可以幫助做出與商品種類優化(確定最佳供應商或推薦人)，每組客戶的獲利目標的定義，確定哪些渠道(或每組中的哪個組)有關的戰略決策。渠道)對我們來說最有利可圖，或者哪種類型的產品最適合該渠道……以及無數的信息，這些信息在業務決策者進行干預并決定根據直覺決定停止決策時才有意義。

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) 😊

需要細分您的客戶？ 馬其頓的菲利普，然后是路易十一世已經說過：“分而治之”……盡管在本文中，報價沒有戰爭意義，但如果您走了這么遠，我希望可以說服您了解客戶的重要性不論您所在的部門或公司的類型如何，數據庫都是組成的，利用這些信息的可能性很多(對于市場和銷售部門而言具有巨大的價值)😊

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.

采取步驟，使戰略和技術部門之間進行溝通……如果您還沒有任何數據，請戴上電池！ 如果您想從使用戰略定位時所提供的選項中受益，請不要再猶豫了，開始收集有關客戶如何成為關鍵客戶的信息。

翻譯自: https://medium.com/ai-in-plain-english/what-is-clustering-customer-segmentation-with-clustering-techniques-cac5125fea8b

抽象類細分舉行

總結

以上是生活随笔為你收集整理的抽象类细分举行_什么是用聚类技术聚类的客户细分的全部內容，希望文章能夠幫你解決所遇到的問題。

如果覺得生活随笔網站內容還不錯，歡迎將生活随笔推薦給好友。