Customers contact companies via multiple channels, from social media platforms and review sites to email and live chats – any time of the day, wherever they are. Add to that the growing number of mobile users making it easier than ever for customers to get in touch, and you can begin to imagine how customer service tickets start to pile up.
So, what happens with all these support tickets? Well, as you probably know, customer agents have to assign tags or categories to each ticket depending on their content, which routes them to the correct team. That in itself takes time because the customer agent has to skim-read every issue before they can tag text with the corresponding category.
Now, let’s imagine a ticket has been incorrectly tagged as System Error and sent to the IT team instead of being tagged as Pricing Inquiry and sent to the sales team. The recipient reads and re-tags the ticket so that it’s redirected to the correct team. Already, ten minutes have gone by just sorting the ticket.
Multiply this situation by 100, and you’ve got a pretty lengthy task ahead of you. Unless, of course, you’ve implemented ticket classification with AI into your customer service.
In this guide, we’ll go into more detail about what ticket classification with machine learning is, how it works, and some of the tools you can use to get started with ticket classification right away. It’s easier to implement than you think, and it can help automate your customer support processes to make your business even more efficient.
Read along or jump to the section of your interest:
Let's start your journey towards customer service automation!
When an issue or support ticket drops into your help desk, first it needs to be processed and assigned a tag or category so that it’s routed to the correct team member. This involves reading the ticket, so that agents know which category to choose.
However, manual classification systems are often complicated and cluttered with too many categories for agents to choose from. After spending endless hours going through tickets, agents will often end up assigning tickets the ‘Other’ category to sort them faster and avoid spending precious time searching for the correct category.
Ticket classification with machine learning avoids this problem, firstly because machine learning tools will only assign a tag that you've previously defined, and secondly because ticket categorization models will assign these tags automatically, without you needing to scroll through a long list of tags.
Instead of humans interpreting content and categorizing it accordingly, automatic ticket classification uses Natural Language Processing, a subfield of machine learning, which helps machines process, understand, and potentially generate human language in a fast and cost-effective way.
Automated ticket classification can be done in various ways, depending on the problem you’re trying to solve. Perhaps you want to improve customer response times by routing your tickets to the correct teams as quickly as possible. Using text classifiers, you can tag tickets as and when they drop into your help desk. For example, you might categorize tickets by language, topic, or specific channel (Twitter, email, live chat, etc), and route them to the correct team member or department based on these tags. Let’s say you have a ticket from a customer who is asking for a refund. A topic classifier would tag this ticket as Refunds, in which case it would be sent to the accounts department. You could also use an urgency detection model, which can flag urgent tickets by analyzing their content for expressions such as right away, immediately or ASAP, and route these to teams that deal with urgent tickets.
We’ll go into more detail about the various ticket categorization models later on. Now, let’s take a look at why ticket classification is important.
Did you know that the average office worker receives around 120 emails every day? That’s a lot of time you need to spend opening, reading, responding or deleting, and tagging. Time that could be better spent on more fulfilling tasks. However, these emails are important, and so is all the other data you receive from various channels, so you can’t just ignore them. To manage all your data without spending hours and hours sorting it manually, you need to use ticket classification tools that will automatically sort your data for you.
Imagine this scenario: customer issues from various channels starts piling up in your help desk, and you rush to tag each ticket manually. Perhaps you’re just about managing your workload, but do you think you’ve been able to process each ticket correctly, and on time? Correct ticket classification is extremely important because this data can help shape your business and give you valuable insights, not to mention streamline processes such as routing tickets to correct team members, and prioritizing tickets that are more urgent.
As companies receive more customer queries in the form of social media posts, live chats, emails, and more, it’s harder for support agents to keep up. That’s why ticket categorization with machine learning is key. Not only is it scalable, it’s also fast and highly accurate if models have been trained correctly. Let’s delve into the benefits of automated ticket classification below:
Rising customer issues, queries, and requests don’t necessarily mean you need to hire more staff. Ticket classification with machine learning automatically tags hundreds of support tickets in seconds, as opposed to hours if done manually by human agents. In other words, you can sort millions of pieces of data at a fraction of the cost of manual methods, save time so that agents can focus on more fulfilling tasks, and avoid inundating teams with heavy and repetitive workloads.
The big advantage of machine learning is that ticket categorization tools can work around the clock. So, whether a customer sends a complaint while on their way to meet friends for dinner, or the IT department of a company in China is having problems upgrading their software, it doesn’t matter what time it is, automated ticket classification can send a response in real-time, 24/7. Plus, if there’s a problem on social media that appears to be escalating, you’ll be able to detect and solve it straight away before it develops into a PR crisis.
As your support tickets, emails, chat conversations, survey responses, and so on, start to increase, you won’t have to worry about diminishing quality. Ticket classification with machine learning enables you to tag your tickets accurately because it applies the same criteria to measure each set of data, plus a machine will never be subjective, lack alertness, and rush through tickets without understanding them properly.
Most customer service teams use help desks to manage their workflows and streamline their processes. However, there are more and more businesses equipping their help desks with AI tools to automate ticket classification, meaning human agents no longer have to sift through endless tickets and tag them manually. Gartner, Inc. predicts that, by 2021, 70% of organizations will assist their employees’ productivity by integrating AI in the workplace.
Zendesk, HelpScout, Freshdesk, Salesforce, for example, are omnichannel support desks that allow you to integrate AI tools, including chatbots, automated ticket classification and routing, and automatic data collection and reporting.
You’ll find various open source tools for text analysis with machine learning, but you’ll need to know how to code before implementing this software into your help desk. However, with AI software, such as MonkeyLearn, you can start classifying your tickets right away.
MonkeyLearn is a text analysis tool that allows you to classify your tickets in various ways. It also offers seamless integrations with numerous help desks, some of which we’ve mentioned above, so you can connect ticket classification models with your apps, quickly and easily, without typing a single line of code.
Sounds too good to be true, right? Well, thanks to MonkeyLearn’s intuitive interface, AI classification tools are really simple to use. Now that you know how accessible machine learning tools are, let’s take a look at some of the ways in which you can classify your tickets with MonkeyLearn.
There are many ways in which you can automate ticket classification, and the one you choose will depend on various factors. TD Bank, for example, is the first bank in Canada to offer customer service via chatbot on Twitter. The customer is able to select from three categories: Customer Service, TD Credit Cards, and Chat with a Specialist, which a chatbot can automatically respond to or route to a specialist.
In customer service, classifying support tickets by topic, sentiment, urgency, and language are the most popular approaches, and MonkeyLearn offers ticket classification models for all four:
By classifying your tickets by topic or theme, you’re able to automate ticket routing in your help desks, so that each ticket is sent to the person best equipped to deal with each issue.
For example, let’s say Amazon receives an issue about a package that says delivered, but the customer never received it. Using MonkeyLearn’s e-commerce support ticket classifier, you can automatically classify this ticket as Shipping Problem as soon as it drops into your help desk. Here’s an example of how MonkeyLearn processes this information:
A topic classifier can also be useful for organizations that have more than one product and different support teams responsible for each one. In this instance, you’d use a topic classification model to categorize tickets by product name (e.g. Google Drive, Google Phone) and then by topic issues that fall under each product (e.g. Login issues, Refunds, etc).
If you decide to classify your tickets by urgency, then you can use MonkeyLearn’s urgency detector. This pre-trained model enables you detect issues that need immediate action, for example social media mentions that contain expressions such as ‘right away, immediately, ASAP’ would be classified as Urgent by MonkeyLearn’s model. You might want to tailor this urgency detector to your business; perhaps by creating a model that recognizes messages from the CEO of a company or premium subscribers. While these might not be urgent issues, you might want to prioritize these tickets over others. Take a look at how ticket classification by urgency works, below:
Ticket classification by sentiment is one of the fastest ways to sort tickets. The pre-built sentiment analysis model available with MonkeyLearn, for example, allows you to classify the polarity of each ticket (e.g. Negative, Neutral, and Positive). In the same way that an urgency detector prioritizes tickets, based on expressions denoting urgency, a sentiment analysis model is also able to prioritize issues, but based on expressions that indicate negativity.
Ticket classification by sentiment can be extremely useful for detecting a serious issue by monitoring ratios of negative to positive tickets over a period of time. Let’s say there’s a sudden increase in negative comments; this might suggest that there’s a serious issue affecting a large group of customers, such as a bug or a video of bad customer service that’s gone viral! Let’s take a peek into how sentiment analysis with MonkeyLearn works:
These ticket classification methods can all be combined to prioritize and route tickets even more effectively. Once you’ve ran a sentiment analysis on a batch of support tickets, for example, you can run a topic analysis on each group of sentiments to find out which topics your customers are talking about negatively or positively. This is called aspect-based sentiment analysis, which you can read more about here.
If you work for an international company, chances are you receive customer issues in various languages. However, instead of agents wasting time bouncing issues from one team to another until they land in front of the correct localized teams, MonkeyLearn’s language classifier is able to route tickets to the right person, right away. Let’s take a look at this example below:
This classifier correctly classified this customer issue as Spanish. Oh, and did we mention that it can recognize 49 different languages!
Once you’ve tested MonkeyLearn’s pre-built models, you’ll be tempted to build your own model to get even more accurate results when classifying your tickets. Luckily, our user interface is really easy to use and, after signing up for free, you can have a go at building your own ticket classifier by following these 6 simple steps:
You’ll need to upload support tickets to train your classification model. You can upload data using a CSV or Excel file. Alternatively, you can import data directly from third-party help desks such as Zendesk and Freshdesk , as well as other apps including Twitter, Gmail, and Front:
As we mentioned earlier on, you’ll need to define a set of tags that are best suited to your organization. Once your classifier has been trained, it will automatically categorize tickets using these tags. Once you’ve defined them, add them to your model:
Finally, you’ll need to tag each example with the expected category to start training the machine learning model:
Once you’ve finished creating your ticket classification model, you’ll need to test it in the ‘Run’ tab. You can do this in two ways, either by choosing ‘demo’ and writing new text directly in the text box field, or by choosing ‘batch’ and uploading new, unseen tickets. This way, you’ll be able to see how accurately your model sorts data:
MonkeyLearn provides some useful tools to determine how well your model is performing in the ‘stats’ section, such as classifier stats (e.g. accuracy, F1 score, precision, and recall) and a keyword cloud for each category:
If you want to improve the accuracy of your classifier, continue training it by clicking the ‘train’ tab. Here, you can tag more data and retag incorrectly labeled examples.
Now that you have trained your classifier, the final step is to integrate it into your customer service software. MonkeyLearn offers some handy integrations such as Zendesk to make it easy to put a model to analyze and tag data in the apps you use every day. Alternatively, if you know how to code, you can integrate the model using the API with your favorite programming language.
Tagging tickets has always been an important role within customer service teams. However, with increasing amounts of customers taking to social media platforms, review sites, and more to complain, it has become harder for support agents to stay on top of all their incoming tickets. While ticket triaging is simple, it’s time-consuming and tedious.
Traditional methods of manually tagging and routing tickets are no longer sustainable, which is why more and more organizations are turning towards ticket classification with machine learning. By equipping their help desks with AI tools, organizations are able to complete simple tasks faster and more accurately and leave more complex tasks to human agents. Automated tools are also scalable, meaning you won’t have to hire extra agents if there’s a sudden spike in tickets.
The best part is that AI-equipped ticket classification tools are readily available, easy to use, and less costly than implementing your own AI solutions. These tools make machine learning accessible to customer service teams and can be easily implemented within your processes on a subscription basis.
MonkeyLearn, for example, can help you get started with AI ticket classification for free, and comes with many pre-built ticket classification models. You can also integrate them into your help desks, as well as tailor them to your company, all without having to write a single line of code. Ready to try out ticket classification for yourself? Request a demo from MonkeyLearn and start automating ticket categorization with AI.
October 3rd, 2019