A never-ending backlog of tickets is all too common in the world of technical support— it’s no wonder why companies are constantly looking to help overworked support teams.  Since there’s no magic wand that can reduce the number of customer issues, companies turn to support automation—often in the form of chatbots— to enable customers to service their own problems.
The issue? Not all support automation is created equal.

According to Gartner, a mere 9% of support journeys resolve solely within the self-service channel. As a result, most support users — even if they start with a self-service solution like a chatbot — end up with your live support team anyway in order to solve their problem.

By trying to solve ticket reduction, chatbots tend to focus on deflection rate instead of solving actual customer issues.

In this post, we will explore the difference between deflection rate and resolution rate, why focusing on the resolution rate is important, and some of the common pitfalls of support chatbots.

What is deflection rate?

In navigating the world of support automation, one buzzword that is commonly discussed is deflection rate.

Deflection rate is the percentage of support requests that are addressed by self-service or self-help tools that would otherwise be serviced by agents. In other words, it refers to the number of tickets your team avoids having to deal with as a result of automation.  

For example, if 23 out of your 100 tickets are intercepted by a chatbot and avoid being sent to your support team, the deflection rate would be 23%.

While deflection rate helps measure tickets not sent to your agents, it ignores the customer experience side of tech support, completely neglecting if customers’ issues were actually resolved. In our example above, 23 tickets deflected does not mean 23 customer issues were fixed.

Where do these 23 deflected tickets go? Good question. These automated tickets can end up in several different states. Ideally, the requests are resolved, and the customer is happy. However, this is seldom the case. Typically these tickets are neglected because of the tickets timing-out, the chatbot failing to provide adequate information, or the user simply becoming frustrated and ending the chat session. All of these negative user experiences are also consolidated into the deflection rate.

Effective self-service automation needs to reduce the volume of issues your team handles and solve customer problems. Without measuring both, companies get a skewed sense of the effectiveness of their self-service support tools, and customers continue to be dissatisfied. That’s where the resolution rate metric comes in.

What is resolution rate?

Resolution rate measures the percentage of support requests that were fully resolved by self-service or support automation. In other words, the resolution rate reflects your actual ability to solve your customers' problems.

Referring back to the previous example, if 23 out of 100 tickets avoid being sent to your customer team, but only 18 were resolved, the deflection rate would still be 23%; however, the resolution rate would be 18%. When considering these tickets, your goal should be to close the loop and resolve your customers’ issues not deflect them. This is why the resolution rate is so important — it indicates that your support loop is completed and your user’s problems are resolved.

So, why should you measure the resolution rate?

Unlike deflection rate, the resolution rate incorporates the customer perspective in the metric. Customer-centric support organizations recognize that incorporating a customer perspective is vital in measuring the success of your support automation as it is when measuring overall customer support. Ultimately, the resolution rate equals a better metric to understand when your ticket loops are closed and accurate insights into your overall customer satisfaction.

Why chatbots talk about deflection rate rather than resolution rate

Chatbots are great at handling requests that have distinct, one-directional logic paths. But what does that mean?

By distinct, we mean that the chatbot can process information with clear and definitive information. No ambiguity or room for interpretation. Think “yes or no” questions, or ones where the answer is obvious, such as “is the light red or green?”

Here's an example of a simple one-directional logic flow that portrays a common conversation users experience when asking for help from a chatbot.

Chatbot one-directional logic - deflection rate

You can see the support path only flows in one direction, and each step needs to have a distinct and clear answer. That works well for more general cases, but there is often more nuance when it comes to product support.

Chatbots typically use deflection as a mechanism to deflect users from the customer support team but not fully solve the user’s problem. As a result, customers will continue to have issues with their product and will either a) restart the troubleshooting process more frustrated than before, potentially taking this out on your support team, or b) give up and move on to a competitor's product. Both end up costing your business money and ultimately affect your companies reputation.

Neither are outcomes that you want. However, the chatbot provider will count this as a success when measuring the deflection rate, simply because the initial request wasn’t sent to your support team.

Chatbots don't want to highlight the times their product doesn't meet expectations; they are incentivized to show off a higher deflection rate because it increases their perceived value. If you don't care about outcomes, of course, having more automated requests is better and leads to less time, effort, and money you have to spend on human support.

Measure metrics that count

Most chatbots aren't very good at actually solving problems, and try to dazzle you with 'deflection' rates. The delta between resolutions and deflections can be high, especially for complex requests like physical product support. Measuring the resolution rate provides clarity into what really matters for the business: how often support automation handles requests correctly.

If you want to help your business and customers, you should focus on 'resolution' rates instead. Unlike a chatbot, Mavenoid is a product expert that helps with the consumer’s needs end-to-end and was created to solve customers’ problems rather than deflect them.

Learn how Mavenoid helps improve your resolution rates here