Glossary · Support metrics

What is deflection rate?

Deflection rate is the share of customer questions handled by self-service, such as a help center or an AI agent, without creating a ticket for your team. Formula: self-service resolutions ÷ (self-service resolutions + tickets created) × 100. 600 self-served questions and 400 tickets give a deflection rate of 60%.

Also calledTicket deflection · Self-service deflection rate

By , pen name of the Asks founderUpdated

Formula

Deflection rate = self-service resolutions ÷ (self-service resolutions + tickets created) × 100

Worked example: In a month, 600 visitors get their answer from help articles or the AI agent and 400 open a ticket. Deflection rate = 600 ÷ 1,000 × 100 = 60%.

The problem with deflection

Deflection counts what didn't reach your team, not what got solved. A customer who reads an article, finds nothing and gives up counts as deflected; so does one who leaves the chat frustrated. That's why the word has a bad name with customers: it can describe a bot that's good at making people go away. Our customer service metrics guide puts raw deflection on its list of numbers to stop reporting alone.

How to make it honest

  • Require a resolution signal: the customer confirmed the answer helped, or didn't come back within a set window.
  • Pair it with quality: CSAT on self-service conversations, and repeat contacts within 7 days.
  • Keep the exit open. Customers who ask for a person should reach one. An escalation path is part of good self-service, not a failure of it.
  • Prefer resolution metrics. For AI agents, AI resolution rate and containment rate say more, once you know how they're defined.

Why the definition affects cost

Several AI vendors charge per resolution, so how they count a deflected or resolved conversation decides the bill. The AI Support Pricing Index lists which tools charge that way, and the AI support cost calculator shows what it means at your volume.

In the product

How Asks handles this

Asks doesn't report a deflection rate. Its analytics show the AI resolution rate, split into confirmed and assumed resolutions, next to conversations transferred to a human and CSAT, so a high automation number can be checked against quality.

7-day free trial · card required · cancel anytime

FAQ

Questions, answered.

01What's the difference between deflection and containment?

Deflection usually describes self-service that kept a question from becoming a ticket, through a help center or a chatbot. Containment comes from contact centers: the share of conversations a bot or phone menu handled without reaching a human agent. Both count the absence of a person, not a solved problem.

02What is a good deflection rate?

It depends on your question mix and how strictly you count. A definition that requires a resolution signal will always come out lower than a loose one, so compare only numbers measured the same way.

How we made this page

Written by Ray (the pen name of the Asks founder) from how support teams use the term and how Asks implements it, with AI-assisted drafting. Formulas and examples are worked arithmetic, not benchmarks; every outside fact links to its source with the date we checked it, and product facts were checked against Asks on 2026-11-17. How we write and correct pages is in our editorial policy.

Answer customers instantly, hand the rest to your team.

Asks learns your website and help center, answers on chat, email and messaging apps, and hands conversations to your team in one inbox when a person should take over.

7-day free trial · card required · cancel anytime

Chat on WhatsApp