Almost exactly a year ago, I published a column titled How to Optimize Zendesk to Deliver Both Effective and Efficient Customer Support. In my third point about the importance of optimizing your Zendesk knowledge base, I left incomplete exploring artificial intelligence as a means of making it easier for customers to find answers to their questions.
In the months following that post, my company upgraded our Zendesk subscription to a plan that included the use of their AnswerBot product. For those that may not know, AnswerBot works in a couple of different ways to help boost customer self-service. It works in both the embedded website chat widget and in the support form in the Help Center, using Natural Language Understanding to present relevant knowledge articles based on customer inquiries. For email support, AnswerBot sends helpful article suggestions in the new ticket autoresponse via email, allowing customers to self-solve their issues without the assistance of a human.
Now, before you go thinking that this is a pro-Zendesk article that’s irrelevant to your customer support team or contact center, I believe that many popular support ticketing systems and knowledge base platforms offer similar technology to boost self-service.
In this article, I’ll talk about my journey with this technology and share four lessons learned while working to improve customer self-service. I hope that you can glean some best practices for your operation.
Understanding the Key Metrics for Self-Service
Once we got access to AnswerBot, I flipped the switch on to see what sort of results we would get. Our subscription includes 500 “free” solved tickets per month. Doing a bit of math, if my cost per contact is around $10, that’s potentially a savings of $60,000 per year in support costs. That’s at least enough to pay for one FTE and our Zendesk subscription for the year — maybe more. It was time to make the most of this.
With 500 solves as the goal, I built a dashboard to monitor our progress on key metrics. The metrics we monitor include:
- Percentage of tickets where customers click on an offered article
- Percentage of tickets where customers indicate that a suggested article solved their issue.
In addition, I try to regularly look to see which articles solve customer issues and which article suggestions customers mark as “not helpful.”
Regardless of what tool you’re using, this insight should be available. Without such reporting, it will be difficult to understand the value and effectiveness of customer self-service.
Lesson #1: Consistently look at reporting to determine the success of customer self-service efforts. Knowing your cost per contact is critical to understanding the true value of this technology.
Continuously improving the content
Start by answering the new, unique questions customers ask
Any initiative to boost customer self-service requires a strong commitment to continuous improvement. Does your knowledge base answer only those questions and answers you thought to ask? Or are you continuously answering questions from actual customers?
I’m convinced it must be a combination of both. Customers constantly ask questions that aren’t answered in our knowledge base, and the real challenge is to empower our customer support team to recognize and write those questions and answers down for the benefit of future customers.
And for good measure, we’ve set goals for customer support team members and are tracking their contributions on performance scorecards that are reviewed in regular one-on-one conversations. Sometimes, as well-meaning support professionals, we find ourselves answering the same questions over and over again. It’s important to think critically and aim to answer questions once for all — whenever possible.
Pay attention to customer feedback
As mentioned earlier, one way we are consistently tuning AnswerBot is to review those tickets where customers indicate that the AnswerBot suggestions were not helpful. My thought process as I review these tickets boils down to this one question:
- Did AnswerBot suggest at least one article that addressed the customer’s question?
If the answer to that question is “No” how can I add a new article or edit an existing one to make it more relevant to the customer’s question? After making updates, Zendesk provides a tool to input the content of the ticket to see if AnswerBot would suggest the new content the next time a customer asks that question.
Now, if at least one article suggestion does address the customer’s question, why didn’t the customer find it helpful? Again, it’s possible that we need to add more information to the article to make it more helpful. In other cases, perhaps the article directs customers to contact customer support. That’s just the thing the customer was hoping to avoid by searching our knowledge base. It may be time to focus some engineering efforts on building more tools so customers can truly self solve their issues.
Lesson #2: You will never be finished improving your self-help content. Make sure that someone on your team is owning this continuous improvement process and make sure they are working to involve your entire customer support team in the effort.
Applying Creativity to an Unintended Consequence
While I spent a lot of time looking at the “not helpful” articles, I spent almost no time reviewing the helpful articles. Why would I?
Well after some time, we began receiving negative customer satisfaction survey responses from customers who had solved their own tickets. What was happening is that the customer would read a suggested article and click the “This article was helpful. Close my ticket” option and then a couple of days later they would receive a survey.
What we began to realize is that, in some cases, the customer found an article to be helpful but they still needed help. After scratching our heads for a bit, we took two actions:
- We triggered a special email autoresponder to the customer letting them know that their ticket was solved. If they still need help, we invite them to simply respond to the email. Many customers have taken us up on that offer.
- We also created an internal alert to quickly review the self-solved tickets so we can proactively reach out to customers if we think there’s more information that would help resolve their issue.
As a result, we eliminated the negative customer satisfaction issue but we’ve also come to realize that AnswerBot wasn’t quite as effective as we previously thought.
For the record, there are also instances where customers quickly find the answers they seek and follow that up with an elated response to the customer satisfaction survey. We certainly take time to celebrate those wins.
Lesson #3: Self-service is only better if it reduces the effort for your customer. Routinely review the customer’s self-service journey from problem to solution. If a customer isn’t happy or their problem isn’t solved at the end of the journey, work to find a better path for them.
Where do we go next?
After several months at this, I’m a bit sad to say that we’ve fallen well short of our lofty goal of 500 self-solved tickets per month. And what success we have had was further diminished by the handful of dissatisfied customers per month.
But I’d be lying if I said I didn’t enjoy the challenge of trying to help customers self-solve more of their issues. The fact of the matter is, it’s fun to make continuous small improvements and watch the metrics improve as a result.
I’ve reflected on this quite a bit and I’m convinced that our next step is to begin identifying the more complex issues customers face while using our service and help them navigate all of the variables to arrive at their desired outcome. And if we can help customers achieve these outcomes with minimal or no human intervention — all the better! Fearing that you may have just interpreted that as me saying that I’m going to unleash a chatbot on our customers, here is my last and final lesson.
Lesson #4: Before you subject your customers to the latest, greatest chatbot — or similar solution — commit lessons 1, 2, and 3 to heart. Your customer self-service will be great because you put in the time and effort to make it great.
And with that, you can be assured that we’ll be continuing on this journey and I hope that I’ll have more lessons to share this time next year — or hopefully sooner.
In the meantime, have you been on a similar path at your company? I’ve done a fair share of figuring things out on my own and would love to hear about some of the lessons you’ve learned along the way — whether you’re using AnswerBot by Zendesk or another tool. Leave a comment below and let’s keep this discussion going.