Glossary

Ticket deflection

Ticket deflection is when a customer gets what they need without a request reaching your support team, usually through self-service content, automated messages or an AI agent. It is a measure of work avoided, not of happy customers, so it should always be read alongside a satisfaction or re-contact measure.

Last updated: 19 November 2026 · By

Fyncall wrote this page. The Zendesk definition links to the source we used, checked 19 November 2026. The worked examples use made-up numbers to show the arithmetic, not real results.

What is ticket deflection?

Ticket deflection means a customer's question is answered before it becomes work for a person on your team. Zendesk's help centre describes it this way: self-service “can result in fewer requests for agents to handle, known as ticket deflection”.

For a Shopify store, the common ways a ticket is deflected are:

  • A shipping email with a working tracking link, so the customer never asks where their order is.
  • A product page or FAQ that answers sizing, materials or delivery times.
  • A chat or messaging conversation that an AI agent or automated flow answers without a person stepping in.

The first two stop a question being asked at all. The third answers a question that was asked. Our guide to reducing Shopify tickets tracks these separately, calling prevention deflection and answering without your team resolution, because they show up differently in your numbers. Some reports give a single deflection rate that mixes both. The percentage formula below counts only conversations that were started, so by the guide's terms it measures resolution.

How do you calculate deflection rate?

There is no single formula, so always check which one a report uses. Zendesk uses a ratio; a percentage of conversations that never needed a person is a simpler alternative.

Zendesk calls its version the self-service score: total user sessions of your help centre divided by total users in tickets. Zendesk gives 4:1 as an example: for every four customers who attempt to solve their own issue with self-service, one submits a request. Note the word “attempt”: it counts tries, not successes.

For conversations handled by automation, a simple percentage works:

Deflection rate = (conversations closed without a team member replying ÷ all conversations) × 100

For example, if 400 conversations start in a month and 220 end without a person replying, the rate is 220 ÷ 400 × 100 = 55%. Use the same period and channels each time.

Is a deflected ticket the same as a resolved one?

No. A deflected ticket only means no person was involved; a resolved ticket means the customer's problem was actually solved. A customer who gives up, cannot find a way to contact you, or gets a wrong answer also counts as deflected.

What happenedCounts as deflected?Was the customer helped?
Tracking email answered the questionYesYes
AI agent gave the correct order statusYesYes
Customer could not find a contact option and leftYesNo
AI gave a wrong date, customer wrote again on InstagramYes, on the first channelNo

The last two rows push the rate up while making things worse. That is why deflection rate on its own is a poor target.

What is good and bad deflection?

Good deflection gives the customer a correct answer without waiting for a person. Bad deflection makes it harder to reach a person, so the question disappears from your numbers without being answered.

Signs of good deflection:

  • The answer comes from live order data or your written policy, not a guess.
  • The customer can ask for a person at any point and gets one.
  • The same customer does not come back about the same issue soon after.

Signs of bad deflection:

  • Contact details hidden behind several pages of help articles.
  • A bot that repeats the same reply when the customer rephrases.
  • Questions that reappear on another channel, such as email after a chat.

To tell the two apart, pair deflection with CSAT or customer effort score, and check how often customers contact you again about the same thing within a window you choose.

How does Fyncall handle deflection?

Fyncall aims for correct answers with an easy way out, rather than the highest possible rate. Its AI agents answer from your Shopify data and knowledge base, and hand over to a person when needed.

  • Escalation triggers. The escalation settings cover a customer asking for a human, negative sentiment, low AI confidence, set topics such as billing disputes, and the same question asked repeatedly. The same page lists escalation rate and escalation reasons as metrics.
  • A queue for handovers. Escalated conversations land in the Needs Agent queue, which any team member can claim from. See human handoff.
  • Draft & Review. The AI writes and your team approves, which lets you check answers before letting a channel run on its own. It is on every plan, including Free.
  • A test sandbox. You can chat with an agent and see its tool traces before customers do.
  • A satisfaction check. The web chat widget can show a post-chat survey with stars, thumbs or NPS.

Fyncall does not charge a fee per AI resolution. It meters AI replies, with an allowance on the Free plan and a larger one on Pro, as shown on the pricing page.

Common questions

What is a good deflection rate?

We do not quote benchmarks we cannot source, and definitions differ too much for one number to be meaningful. Track your own rate over time, measured the same way, next to CSAT.

Does a knowledge base count as deflection?

A customer-facing help centre does, when people read it instead of writing in. Fyncall's knowledge base is a library the AI agent answers from, so its effect shows up as conversations the AI resolves.

Should I hide my contact email to raise deflection?

No. It raises the number without solving anything, and the questions can come back elsewhere, from customers who are now more frustrated.

Ready to try Fyncall?

Install it tonight, let the AI draft your first replies, and switch it on one channel at a time as you start to trust it.

  • Free plan, no credit card
  • 14-day free trial on Pro
  • Unlimited human replies on every plan