Demystifying Dewu Customer Service IM Full-Link Communication Process

Introduction: From the user's point of view, the process of consulting customer service that we see is robot -> transfer to manual -> evaluation . Human conversation is the core of IM communication in the whole process. What process will the entire customer service system go through before the user touches the manual session? How does the customer service staff ensure that the user can accurately and quickly match the user's doubts and demands, let's find out.

Let's first look at the overall architecture diagram of the customer service session and docking system, as shown below:

• Visitor terminal: user touch terminal (c terminal), with three main functions: manual session, satisfaction evaluation, and robot session
• Customer service terminal: The main docking platforms are: Octopus one-stop workbench (manual conversation chat), robot management background (robot dialogue related), work order workbench (closely connected with manual conversation), online customer service workbench (for artificial conversation) Serve).

In order for the customer service to be able to connect to the user and accurately assign it to the designated customer service, we have developed an online customer service workbench, which has set a series of rules for service time setting, human-computer interaction, incoming line allocation, and queuing overflow conversation flow.

1. Start with the service time setting

1.1 Dewu customer service hours

First of all: the regular service hours of Dewu. Only within the limited service interval, manual conversation consultation can be conducted. For consultation beyond the regular service hours, robot Q&A or message can be conducted. After the manual customer service goes online, the response can be answered.

Regular service hours: weekdays and weekends from 8:00 a.m. to the next morning

Unconventional service hours: Of course, in addition to regular hours, there are special service time settings. For example, during the Double Eleven, 618 and other big promotions, the user traffic increases, and the order volume increases. In a certain period of time, the service time needs to be advanced or extended, and special settings can be made. . In line with the special time entry time, you can also enter the line normally to the manual customer service.

2. Diversion rules

As everyone who has experience in online shopping and has conducted customer service consultations knows, the questions you want to consult include questions before placing an order, questions after placing an order, questions about the logistics progress , and consultation on the size and color of the arriving goods .

Not only these, there are also targeted problems when buying different products, such as: footwear, cosmetics, clothing, bags, watches, 3C categories and so on. These questions need to be answered by professional customer service, so it is not possible to assign a customer service to wiring, so before users enter the line for consultation, it is necessary to match the line shunting rules , which is a complex logic.

Users come from different channels. From the configuration dimension, first make a distinction by channel:

• What are the incoming channels for Dewu?

At present, there are many incoming channels for Dewu, including Dewu App and Mini Programs.

• How to distribute incoming lines according to channels?

For example: For example, enter the online customer service from the product details page of the Dewu App

According to the channel (Dewu App) + source (product details) + diversion dimension + distribution method, it can be assigned to the designated customer service group. (The diversion dimension can set robot priority rules)

3. Queue overflow rules

As mentioned above, it meets the line entry time and the channel source allocation. Finally, it is time to enter the line to manual chat, but it is a little late. Suppose there are only 10 customer service personnel, and the simultaneous connection volume of each customer service is 10. It happens that you What happened to the 101st one. A queuing rule was triggered accidentally.

3.1 Queue process

In order to prevent the customer service from connecting to the queued users who have left and waiting, it is first necessary to determine whether the user is on the customer service page. If the user is not online, the user will be prompted that the push queue is about to end, and the online status will normally enter the queuing process.

Schematic diagram of queuing on the App side:

3.2 Overflow message

When a large number of inquiries are queued during the peak of customer service business, the queuing volume will be uneven and cannot be controlled. In order to ensure that the queue volume of each group and the waiting time of users are balanced, the overflow rules are used to reasonably allocate human customer service resources.

3.3 Real-time data monitoring

There is monitoring data on the monitoring large screen to reserve the number of queued people, and the number of people entering and leaving the reserved queue is only counted once, but if the user enters the line multiple times and queues up multiple times, it will be counted multiple times.

Real-time data: data for the day. View reserved queue times for different queues according to the selected customer service group

According to the monitoring of the number of people in queue and the length of the queue, it is judged whether the user behavior and customer service connection capacity are in line with the current number of users

4. Incoming manual customer service

After the previous series of rule matching and incoming line distribution, we finally came to the manual conversation. At the stage of manual customer service chat, how do customers and our customer service interact with each other? Students who have used the Dewu App and consulted the manual customer service should all know the usage of the client, so how does the customer service connect and communicate with the user? We developed two generations of systems before and after.

4.1 Online customer service system

This system is the first generation system of IM customer service system. Its functions include not only online chat, but also the configuration and user management related to chat are managed here in a unified manner.

4.2 Octopus One-Stop Workbench

That is, the one-stop workbench used by customer service now. Because customer service chats are not only accurate and fast, but also complex session switching information caching, etc., and configuration-related issues will affect the use of customer service, and then developed the Octopus one-stop workbench for customer service wiring-related functions.

The form of a three-column layout:

Left area: online, offline, message user list

• Middle chat area: conversation content, customer service chat area, including functions for sending expressions, pictures and videos, etc.
• Right area: User-related information functional area, including user information, prophets, order query, work order creation (the work order data will be synchronized in the work order system to facilitate special processing of the work order), and product recommendations.

4.2.1 The process of data reaching both ends

The data mainly includes ordinary text messages, rich text, pictures, videos, files and so on. It is a good choice to ensure that messages can be delivered to customer service and client long links accurately and in real time. These data are sent to the backend first - the backend is pushed to the gateway - the gateway is pushed to the Octopus workbench through the long chain , and the current long link is available. Under the hood, each request is simply data sending and receiving.

At this point, you may have questions, what should I do if there are problems such as network jitter in the process of sending messages, so that the problem cannot be reached?

At present, the ACK three-time reconnection mechanism used by the customer service system will retry if it does not receive a response from the gateway within a certain period of time, and perform three reconnections to ensure that the message can be sent successfully when there are occasional unexpected network conditions. At the same time, in order to avoid continuous retry, if the network is blocked by a large area of retries and still fails after three attempts, the sending failure will be displayed. The news does not fall into the library.

4.2.2 SOP for Agent Auxiliary Capability

SOP is the capitalization of the first letter of the three words "Standard Operating Procedure", that is, standard operating procedure, which refers to the description of the standard operating procedures and requirements of an event in a unified format, which is used to guide and standardize daily work.

For the chat process, there are many repeated questions. How does the customer service solve the user's problem faster? In the early days, we have accumulated and improved many knowledge documents for customer service instructions, of which 900+ questions need to be mastered. In order to improve the service efficiency of the customer service side, the agent assistant tool is extended.

According to the information content in the chat, the corresponding knowledge content is automatically matched, and then the customer service determines whether it conforms to the current chat scene according to the matching content, and it can be sent to the user if it matches. As much as possible, the business logic that is streamlined and frequently concerned by customer service is incorporated into the sop-based "smart customer service agent assistant", providing a unified one-stop query, decision-making, and operation portal (tool), simplifying customer service operations and improving customer service. Reception efficiency.

5. End the session

5.1 The session process has several stages

5.2 Session close

There are several ways to close a session

• When the customer service does not receive the user's message within a certain period of time after replying to the user, the countdown to the end of the session will be performed first, and the push session will end after XX. If the user does not reply before the end of the countdown, the default closing session logic is triggered, the conclusion is pushed, and the session is closed.
• When it is confirmed that the user's question has been answered and there are no more questions to inquire, the concluding remark is pushed and the session ends.
• The user has been answered, and there are no more questions to ask to actively end the session.

5.3 Invitation to comment

Before ending the session, customer service can send an invitation to comment to investigate user satisfaction . After the session ends, if there is no invitation to comment, the system will automatically push the invitation to comment. During the session, the user can also initiate an evaluation to give feedback on the satisfaction of the session. In order not to excessively invite comments and reduce the user experience , there is only one opportunity to invite comments per session , and no comment invitation information will be sent no matter whether users have comments or not.

6. Summary

The evolution of customer service has been from the initial platform to the mature system. With the joint efforts of the entire Dewu customer service team, and continuous improvement and optimization step by step, a system that is still stable and easy to use under the pressure of huge number of sessions .

The above is an overview of some main contents in the IM communication link of Dewu. After reading this, you have a preliminary understanding of the operating environment and process of customer service. In addition to these contents, there are many detailed technical contents, welcome to leave a message for discussion.

Text/Jun

@德物科技public account