Using Chatbots to Improve CRM Data: a WeChat Case Study

One-to-many social media is great for reaching a wide audience but struggles to make the communication personal and sometimes even relevant for individual customers. Chatbots and simple automated responses are a way to make personalised communication scalable within messaging apps. However, there is more to chatbots than simply communication. If designed and implemented well, a chatbot can be a vital tool in data gathering, helping companies better understand their customers.

The Rise of Messaging Apps

One-to-many social media platforms (Twitter, Weibo, Facebook) have been losing users’ screen time to messaging apps over the past few years. Messaging is the fastest growing social behaviours online, and messaging apps are consistently in the top downloads in most countries.

As messenger apps have sought routes to monetisation, courting businesses through advertising (as traditional social media has done) was only one path. Innovative messenger apps such as WeChat, Line or Kaokao Talk offer users other services including online and offline payments, games, commerce, and access to various forms of media, on top of the services one would expect of an advanced messaging app, calls, texts, image & video sharing etc.

The issue for marketers is that one-to-one communication on messenger apps is simply impractical with a large number of customers. As social ecommerce grows, scalable personalised conversations needed to be made available to companies.

Many messenger apps offer tools to make this kind of communication possible. From simple keyword-response tools to complicated API integrations, messenger apps offer businesses some interesting functions to design personalised communications, though it is still early days for such integrations.

What’s a Chatbot?

A chatbot is a program, or series of programmed responses, designed to simulate (as best as possible) a conversation with a human. Chatbots can be thought of as another way for users to interface with a service online, whether booking a taxi, ordering a delivery pizza or checking tomorrow’s weather.

Chatbots within messenger apps tend to be text-based and range from offering the user simple button or keyword triggered responses through to a natural language artificial intelligence (AI).

Natural language AI chatbots attempt to “understand” a user’s question or statement and respond accordingly, often AI is experimental and expensive, however Siri and the Google personal assistant are great examples of this kind of interface in action.

Simple keyword or button or keyword response chatbots are far more common, due to the ease in development, and rely on the user only using a limited set of words or phrases to trigger responses from the bot. It is this kind of simple chatbot that we will look at now.

Case study: Meici on WeChat

Online fashion flash sale site Meici (美西时尚) runs a WeChat account that takes advantage of WeChat’s keyword response tool (a very basic chatbot) to gather information about its user-base. When a new user first follows the account, a welcome message instructs them on how to trigger responses.

The message basically reads, “Welcome to Meici. As China’s leading luxury shopping platform, we will strive to provide you with the latest information and content about our range of products.” It then goes on to list 4 ways in which users can interact with the account.

  1. Type “what should I wear today”, “seeking styling” or “looking for what works together” to receive advice
  2. Type in a brand name such as “Gucci” or “Balenciaga” to receive a product list
  3. Enter a category such as “bags” or “high heel shoes” to receive a product list
  4. Type in “Meici is the easy way to enjoy luxury” to receive a 100RMB voucher to spend

When a user first interacts with the account, for instance typing, “what should I wear today”, they receive a message asking if they are a “handsome male” or “stylish female”, upon replying to this the user receives a message containing links to a full outfit. Any time the user then writes “what should I wear today” they instantly receive a gendered new outfit choice message, the system has remembered their gender.

Simple map outlining Meici’s autoresponses

This is a really simple version of a chatbot. However, if the data is being stored and used correctly, it’s a powerful tool for the business. Knowing how many of your engaged user-base is male or female, what brands they’re interested in and what product categories they want to browse, will allow marketers to better tailor content, buyers to better predict needs and businesses to take better advantage of trends.

Designing Chatbots

There are some great examples of more sophisticated chatbots than the one outlined about. Bots are already in use on many different platforms by companies such as airlines, hotels and a wide number of online services.

Any company looking to communicate with customers and clients on a personal level, but at a large scale, should take interest in what chatbots can do. However, just thinking of chatbots as a communication device is missing an important potential application for the tool. As the above example shows, a chatbot doesn’t need to be complex to gather useful data, if it is well designed with both users and business needs in mind.

When planning and designing chatbots, ensuring that the data collected is going to be useful, and will not seem invasive to users is a difficult balancing act. However, getting this balance right can have great implications for building effective CRM data through messenger apps.


This story was first published at Half A World, visit us to keep up with the latest developments in digital in the Asia-Pacific

If you enjoyed this article, help others discover us by hitting the little green heart below.