Containment rate
The containment rate is the share of customer conversations that stay in self-service and never reach a person on the team. The metric comes from call centres and is now used for chatbots and AI agents. It is easy to flatter, because a customer who gives up in frustration also counts as contained.
What is the containment rate?
Containment means the conversation stays in the automated channel and does not land with the team. The metric comes from call centres with phone menus, where the question was how many callers got by without an agent. Today it appears in nearly every report on chatbots and AI agents.
How is it calculated?
Conversations without a handover to a person, divided by all self-service conversations, times 100. If a chatbot has 1,000 conversations and passes 200 to the team, the containment rate is 80 %.
Why can the containment rate mislead?
The formula does not ask whether the customer was helped. It only asks whether they reached a person. So conversations count as contained when the customer
- closes the window after the third unhelpful answer,
- finds no way to reach a person and gives up,
- tries again an hour later by phone or email.
An example: Mr Baker asks a furniture store's live chat whether his sofa will arrive before Christmas. The bot points to the general delivery times twice. Mr Baker closes the chat and calls. The report shows a contained conversation, the phone line gets an annoyed customer. A high containment rate can therefore also mean that handovers do not work well.
Containment rate or automation rate?
The automation rate only counts conversations the AI actually resolved. It is the more honest metric. If you still use the containment rate, back it up with other figures:
- Customer satisfaction after the conversation, for example via CSAT.
- Repeat contact: does the same customer get in touch again shortly after, on the same or another channel?
- Drop-offs: how many conversations end without the customer replying to the last message?
How do you keep an eye on it in SendSeven?
With an AI agent that hands over in time, your analytics show separately how many conversations an agent resolved and how many it handed over to your team. Because the agent hands over as soon as a customer asks for a person, frustration ends up with your team instead of nowhere. All channels come together in one shared inbox.