Wouldn’t it be nice to accurately predict the satisfaction of your friends, family, and coworkers? Think about all the bad gifts that could be avoided. The awkward jokes that could be left unsaid. Well, we haven’t worked out the science on misguided anniversary presents, but Zendesk has made a tool that can predict the satisfaction of your customers.
Satisfaction Prediction is our powerful new tool that helps your support team anticipate and prevent unhappy customers. At its core, Satisfaction Prediction is just a score from 0-100 for each ticket in your system describing the predicted likelihood that the customer will respond with a "Good" satisfaction rating once the ticket is solved.
But there are many beneficial ways you might use the score:
- Add it as a column in your Views
- Use it to prioritize tickets within Views
- Create new Views to highlight tickets in a certain range of values
- Display the prediction score in each ticket’s Events thread, so you can see how the prediction changes with each interaction
You could also Add it as a criteria to Triggers and/or Automations:
- Notify someone when a ticket is in trouble
- Increase priority when a ticket is in trouble
- Route to agent groups based on the prediction value
- Tag the ticket so it appears in a certain View
Satisfaction prediction has the potential to change the way your team processes customer support requests. The positive outcome is that it can empower your team to deliver better support to your customers. But there’s something important to consider when you’re looking to roll it out. As it turns out, being able to predict satisfaction is not without its psychological hangups. The cognitive weight of the prediction score might affect how and when agents address certain tickets. And from a manager’s perspective, using the prediction score can significantly change your support team processes. Let’s explore those two considerations.
Never Tell Me The Odds!
Knowing the prediction score can have a psychological effect on your agents. Experienced or empathetic agents can develop a good intuitive feel for the customer’s mood, and a prediction value that disagrees with that can be confusing or frustrating. On the other hand, seeing the score calculated after each interaction can make the agent feel self-conscious. They might second-guess themselves and become overly focused on the possibility of a bad outcome, instead of focusing on the customer’s need.
As a result, agents might “cherry-pick” tickets to avoid customer interactions that seem like they’ll end badly. Alternately, if an agent sees a high prediction score on a ticket, they might take the customer for granted and not address that ticket right away. This could lead to a worse customer experience.
To help prevent these issues, educate your team about how the prediction scores are calculated, so they have a better understanding of how it functions. It’s also possible to limit who can see the prediction value on tickets by restricting visibility to specific agent roles. You can still use satisfaction prediction to influence how tickets are prioritized or routed, without showing your agents the prediction value itself.
When daily work processes change, a little anxiety is normal. If an agent has mastered the old way of doing things, they might feel like they’ve wasted the that effort and are losing their “expert” status in the team. A job that has become satisfyingly routine can become difficult and frustrating, at least for a time, as they change their habits to conform to the new process.
If you’re using Satisfaction Prediction to change the way tickets are prioritized away from a simple “first come, first served” method, your agents might feel that patient customers are deprioritized in favor of “squeaky wheels”. This feeling comes from genuine empathy for your customers—something you don’t want to discourage!
But the balance can be tricky. For instance, if you’re considering notifying support leaders on tickets that have low prediction values, consider how that will affect their time. Their day will be interrupted to evaluate the ticket and provide guidance to the agent, or craft a response themselves. It’s probably not useful to send them notifications if they don’t have the time to take substantive action.
If you ignore these potential problems and roll out your changes without making the case for the new system’s benefits, you could create morale and even productivity issues in your team. However enthusiastic you are about the positive changes, you don’t want to leave your team feeling you’ve brought them someplace they don’t want to be. Like Bieber at this basketball game:
Use the force (of change management)
Rolling out any new process, based on Satisfaction Prediction or other features, requires successful change management. Here are some suggestions about how to use change management techniques to help ensure a smooth rollout of satisfaction prediction:
Quantify the need: Identify past tickets that could have been helped by satisfaction prediction (such as bad Customer Satisfaction (CSAT) received because slow response or unresponsiveness to customer language/tone), and what the potential upside might be from improvement (% increase in CSAT average, NPS increase, churn avoided, sales increases, greater productivity because happier customers require less time & effort to handle, etc.).
Gather your champions: Assemble a team of leaders within your support organization (managers, team leads, experienced agents) to articulate a vision of how the customer and team experience might be improved by the adoption of this feature.
Plan your strategy: Have your team build one or more plan options for how you might adopt this feature. Which (if any) business rules would change - Views, Triggers, Automations? Would tickets be prioritized differently? Would leaders be notified in certain circumstances? Would team processes change, e.g. in handling tickets identified as having a high probability of dissatisfaction?
Test your hypotheses: Have a small team try the new experimental business rules and/or processes for a defined period of time. Keep the rest of the team informed with what they’re doing (especially important if you’re enabling the display of predicted CSAT in tickets). One way of doing this would be to create a temporary Group for the test agents, then assign a subset of tickets to them using a trigger (e.g. tickets of a certain issue type) that tags those tickets so the new/revised business rules only operate on them. If you have development resources, you might instead use our API to create “dummy” tickets based on actual customer tickets, perhaps having a “red team” of agents that acts as the customer.
Evaluate your test: Did the changes help the customer, and/or the agents? Are there tickets that would have had better outcomes with your older process? It the test didn’t meet your expectations, it might be worth making revisions and trying the test again, perhaps with a different team. Using a different team means you might have individuals that react differently, so the tests might not be equivalent, but be involving more of the team, you increase the opportunity for buy-in from the team as a whole. The more people who can vouch for the change, the more likely everyone else will be willing to jump on board.
Share your findings: Once you’re satisfied with the new processes, build a presentation or training (depending on how extensive the changes are) and deliver that to your team.
These are good strategies to use when rolling out any major changes to your process. Following these guidelines can help keep your team engaged and, dare we say, satisfied.
Head to the forums for more information about Satisfaction Prediction
Keep the knowledge flowing! For more tips like this, check out Zendesk's Tip of the Week collection.