Fallback Response
Definition
A fallback response is the chatbot's safety net for situations where normal processing fails — whether due to low NLU confidence, out-of-scope queries, or missing knowledge base content. A well-crafted fallback response acknowledges that the bot did not understand or cannot help, apologizes briefly, and provides a constructive next step: ask the user to rephrase, suggest related topics, link to relevant resources, or offer escalation to a human agent. Fallbacks are triggered by confidence thresholds in the NLU layer or by explicit checks in the dialogue flow.
Why It Matters
How a chatbot handles failure is as important as how it handles success. A poor fallback with no next step leaves the user stuck and frustrated. A good fallback maintains trust, keeps the user engaged, and often resolves the situation through clarification or escalation. Monitoring fallback rates is also one of the most valuable signals for identifying gaps in the bot's knowledge or NLU coverage.
How It Works
When NLU confidence scores fall below a defined threshold, or when no knowledge base result meets a relevance cutoff, the dialogue manager routes to the fallback handler. The fallback handler selects an appropriate response — which may vary based on how many fallbacks have occurred in the same session. Many systems implement a tiered fallback: first ask for rephrasing, then offer related FAQs, then escalate to a human.
Real-World Example
A user asks the bot about options trading strategy — completely outside the bot's scope for a software company. The fallback triggers: 'I am not sure I understood that. I am here to help with questions about our product. You can try rephrasing, or speak to our support team.' The user is not stranded.
Common Mistakes
- ✕Using a single, repetitive fallback message that users see every time, eroding trust.
- ✕Triggering fallback too aggressively — setting confidence thresholds too high causes the bot to fall back on queries it could have handled.
- ✕Not logging fallback events for analysis, missing the opportunity to expand the bot's capabilities based on real gaps.
Related Terms
Escalation to Human
Escalation to human is the process by which a chatbot transfers a conversation to a live human agent when it cannot resolve the user's issue. Effective escalation passes the full conversation context to the agent, ensuring the user doesn't have to repeat themselves and the agent can immediately continue where the bot left off.
Intent Recognition
Intent recognition is the process by which a chatbot identifies the goal or purpose behind a user's message. It classifies free-form user input into predefined categories (intents) — such as 'check order status', 'request refund', or 'get pricing' — enabling the bot to route the conversation appropriately.
Chatbot Analytics
Chatbot analytics is the measurement and analysis of chatbot performance — tracking metrics like conversation volume, resolution rate, fallback rate, escalation rate, and user satisfaction. These insights reveal how well the bot is performing and where to focus improvement efforts.
Conversation Logging
Conversation logging is the practice of recording and storing chatbot conversation transcripts for analysis, quality assurance, compliance, and training purposes. Logs capture every message exchanged, enabling teams to review interactions, identify failures, and continuously improve the bot's performance.
Chatbot Testing
Chatbot testing is the process of evaluating a chatbot's performance before and after deployment — verifying that intents are correctly recognized, flows execute as designed, edge cases are handled gracefully, and responses meet quality standards. Regular testing prevents regressions and ensures the bot delivers a reliable user experience.
Ready to build your AI chatbot?
Put these concepts into practice with 99helpers — no code required.
Start free trial →