Sentiment and frustration

The worst outcome in automated service is a system arguing politely with a customer who is upset. The agent notices how a customer is feeling as the conversation happens, changes its approach when frustration or distress appears, and brings in a person under the rules you set. After the conversation, the same reading is recorded against it, so satisfaction is measured across every conversation, not just the few customers who answer a survey.

These are signals for your team to act on, read from what the customer says and how they say it. They are not certainties about what anyone feels.

During the conversation

The agent assesses the customer's tone and language throughout the conversation, not just at the start. On voice, that includes how something is said as well as the words. On messaging and email, it reads the words, the way they are written and how the conversation is going: a customer who repeats the same question three times is telling you something, even if every message is polite.

When the reading changes, the agent changes its approach. It acknowledges the problem before giving options, drops anything that is not essential, and stops offering alternatives the customer has already turned down. See Conversation design by situation.

When frustration leads to a person

Frustration is one of the triggers your team can attach to escalation rules. When the agent detects frustration or distress that meets the rule for that workflow, it hands the conversation to the person or queue you have named, with the full conversation and a brief. The customer does not start again. See Escalation rules and Handover to your team.

You decide how sensitive each trigger is, per workflow. A property developer may hand over any buyer who sounds frustrated about a handover delay at the first sign. A hotel group may let the agent finish a simple booking change first, unless the guest asks for a person.

After the conversation

When a conversation ends, a sentiment reading is recorded against it: how the customer felt about the interaction as a whole, and how that changed from the start of the conversation to the end. A conversation that started frustrated and ended settled is recorded differently from one that started settled and ended frustrated, and that difference is often the most useful thing a quality team can see.

Where you want more than positive and negative, emotion analysis is configured during implementation. It distinguishes, for example, a customer who is frustrated from one who is confused or anxious. The distinction matters because each needs a different response: a frustrated customer needs the problem acknowledged, a confused one needs a clearer explanation, and an anxious one needs to know what happens next.

Satisfaction from every conversation

Satisfaction surveys reach the customers willing to fill them in, usually the most pleased and the most annoyed. Satisfaction read from the conversation itself covers every customer who got in touch, on every channel. It is reported by channel, by reason for contact, by department and over time, alongside outcomes and quality scores. See Outcomes and reasons for contact.

The two approaches work well together. If you run a survey, the conversation-based reading shows how the customers who did not answer felt, and the survey shows what the customers who did answer chose to say.

Reading the signals well

Sentiment is read from language and tone, and people express themselves differently. Some customers are brief when content and others are brief when angry. Gulf Arabic, English and a mix of the two carry feeling in different ways.

The readings are built and tested on conversations in the languages and dialects your customers use, and your quality team reviews them during calibration. See Quality review on every conversation.

Treat a sentiment reading as a reason to look, not a conclusion. Use it to find the conversations worth opening, to see which reasons for contact frustrate customers most, and to set escalation rules. Do not use it on its own to judge an individual customer or colleague.

In practice

A travel group uses the agent for booking changes on voice and WhatsApp during a period of flight delays.

1. Its escalation rule for delay conversations says: hand over when frustration is detected twice, or at once if the customer mentions a child, a medical need or a missed connection.
2. Faisal calls about a delayed flight. He is calm while the agent confirms his new departure. He becomes frustrated when he learns his connection will be missed.
3. The missed connection triggers the rule. The agent tells him it is passing him to the team who can rebook his connection, and transfers the call with a brief.
4. After the call, the conversation is recorded as handed to the team, with a sentiment reading that moved from settled to frustrated.
5. At the end of the week, the customer experience lead sees that delay conversations mentioning missed connections produce the most frustration, and adds a rebooking option for connections to the agent's permitted actions.

What your team controls

• Which workflows use frustration as an escalation trigger, and how sensitive each trigger is.
• Where each handover goes, and what the brief contains.
• The signals that route a potentially vulnerable customer to a person.
• Whether emotion analysis is configured, and for which workflows.
• Who can see sentiment and satisfaction readings, by role.

Related

• Escalation rulesGovernance and control
• Handover to your teamAgents
• Outcomes and reasons for contactQuality and analytics
• Insights and trendsQuality and analytics
• Conversation design by situationAgents

Product names and logos are trademarks of their owners. Their mention shows systems Telonic connects to and does not imply partnership or endorsement.