Deflection Rate vs Containment Rate
Deflection rate is the share of all support demand resolved without a human agent, including pure self-service where no conversation ever started. Containment rate is narrower: the share of automated conversations a bot handled from start to finish without transferring to a human. Every contained conversation is a deflection, but not every deflection is a contained conversation.
The shared blind spot
Both metrics reward the customer not reaching a human, and neither one, on its own, proves the problem was solved. A bot can contain a conversation by frustrating the customer until they abandon it, which counts as containment and as deflection while being a support failure. Read alone, both numbers can reward exactly the wrong behavior.
What to measure instead
Use deflection and containment as operational signals, but judge success on verified resolution and reopen rate. If containment is high but reopen rate is climbing, the bot is trapping people, not helping them. The outcome metric always outranks the avoidance metric.
How Resolve handles it
Resolve treats containment as a means, not an end. Reva is built to contain by resolving, not by stalling, so containment is always read alongside auto-resolution and CSAT. When Reva cannot resolve within its confidence threshold, it escalates cleanly with a decision card rather than looping the customer, so a contained conversation is one that was genuinely solved.