Containment rate (containment rate/self-processing conversation) is the percentage of conversations that the chatbot processes completely without needing to transfer people. For example: 100 conversations, chatbot handles 70 → containment 70%. This index shows how much work the chatbot can "carry" - high is not the final goal, accompanied by CSAT is meaningful. Zenify recorded 98.4% AI accuracy (approved).
TL;DR
- Containment = % of chatbot conversations are completely processed by themselves.
- Calculate: number of bot conversations processed ÷ total conversations × 100.
- High + good CSAT = effective chatbot.
- Increase by: adding knowledge base, script, fallback.
- Don't chase away containers absolutely — difficult customers still need people.
What is Containment rate?
Containment rate is the percentage of conversations that are fully handled by chatbots — customers are answered and finished without human intervention.
Terminology | Meaning
Contained conversation | Chatbot handles complete self-processing
Escalated conversation | Switch to agent
Containment rate | % of conversations that can be handled by themselves
Calculation method
`
Containment rate = (Number of self-processed chatbot conversations ÷ Total conversations) × 100%
`
For example: there are 1,000 conversations in a month, 750 chatbots handle themselves, 250 people are transferred → containment 75%.
Vì sao containment quan trọng?
Cost savings: Bot conversations handle themselves without requiring an agent.
Speed: bot responds instantly, customers don't wait.
Uniform quality: the bot answers correctly according to the knowledge base every time.
Agent focuses on difficult tasks: Only complex tasks go to people (see Chatbot replaces employees).
How much Containment is good?
There is no absolute "standard" number — it depends on the industry and the complexity of the question:
New businesses have chatbots: 30–50%.
Chatbot is well built: 50–70%.
Simple questions: maybe more than 70%.
More importantly: Containment increases over time and CSAT does not decrease — meaning it is improving in the right direction.
How to increase containment rate
Add knowledge base: Which sentence has the most fallback → add it (see Knowledge base chatbot).
Optimize the script: Where does the customer stop → edit the script.
Improve fallback: don't move people too early — but don't keep customers stuck either (see Fallback chatbot).
Intent recognition training: add different ways of asking for the same need.
Measure regularly: monitor weekly, detect decline early.
Warning: do not evict the container at all costs
If the chatbot "holds" difficult conversations, gives wrong answers → customers are upset, CSAT drops.
High Containment but low CSAT = chatbot makes more mistakes than serves.
Always measure containerment with CSAT and people transfer rate (see CSAT).