What it means
Containment rate measures how well your AI and self-service channels hold customer questions end-to-end. A high containment rate means fewer customers are finding the AI inadequate and demanding human escalation.
Containment rate = (conversations resolved entirely within the AI or self-service channel ÷ total conversations started in that channel) × 100. The key distinction from deflection rate is scope: containment rate applies only to conversations that were initiated in a specific channel (usually AI chat or IVR), whereas deflection rate is a broader measure of all demand kept away from human agents. In ecommerce, containment rate is most meaningful for AI chat widgets deployed on product pages, checkout pages, or post-purchase tracking pages. If 100 customers start a chat session and 75 get a complete answer without asking to speak to a human or leaving unsatisfied, the containment rate is 75%. Containment rate below 50% typically signals one of three problems: the AI can't answer the most common questions (knowledge gap), customers don't trust AI answers and instinctively demand a human (trust/tone gap), or the AI is too quick to escalate on its own (threshold calibration issue). Benchmarks vary by use case: order-status bots typically achieve 80–90% containment; general-purpose ecommerce support bots 55–75%; complex product advisory bots 40–60%.
Why it matters
Every conversation that escapes containment adds to human ticket queue. High containment rate is the most direct measure of an AI channel's effectiveness as a self-sufficient support touchpoint. It's also a customer experience signal — customers who reach the AI, get their answer, and leave satisfied are experiencing exactly what good AI support looks like.
How Bookbag helps
High-accuracy AI that earns trust
Bookbag's answers are grounded in your actual store data and policies — not generic responses — so customers trust the AI's answers and don't reflexively demand a human.
Containment rate analytics
Track containment rate per channel and per intent so you can see which question types the AI handles well and where gaps are driving customers to escalate.
Smart escalation (not premature)
Bookbag escalates when the AI genuinely can't help — not as a default. Calibrated confidence thresholds keep containment rate high without risking wrong answers.
Frequently Asked Questions
See Bookbag in action
Join the ecommerce teams resolving more tickets, answering 24/7, and turning support into a revenue channel with Bookbag.