What Is Ticket Deflection? Definition, Formula, and Benchmarks (2026)
Ticket deflection is the share of support requests resolved without a human agent. Here is the plain definition, the formula, realistic benchmarks, and how to turn deflection into measurable savings.
Ticket deflection is the share of inbound support requests that get resolved without a human agent. A customer finds the answer in a help article, a chatbot, or an AI agent and never opens a ticket for a person to handle. You measure it as deflected requests divided by total inbound requests, expressed as a percentage.
That is the whole idea. Everything else, the formula, the benchmarks, the dollars, follows from that one sentence.
Jump to a section:
- The ticket deflection formula
- What counts as a deflection
- What is a good deflection rate
- How deflection turns into savings
- Deflection vs avoidance
- How to actually raise your deflection rate
The ticket deflection formula
The math is simple:
Ticket deflection rate = (requests resolved without an agent ÷ total inbound requests) × 100
If 2,000 people reach out in a month and 900 of them get a complete answer without an agent ever stepping in, your deflection rate is 45%.
The number that matters for budgeting is not the percentage, though. It is the count. Nine hundred deflected tickets, each one an agent reply you did not have to staff and pay for. To put a dollar figure on it, plug your real volume into the ticket deflection calculator and it will net out the savings after the cost of the automation.
What counts as a deflection
A deflection only counts when the customer's question is actually resolved without a human. That distinction matters, because a few things masquerade as deflection and are not.
Counts as deflection:
- A help-center article answers the question and the customer leaves satisfied.
- An AI agent resolves the conversation end to end, no handoff.
- A self-service flow (reset password, track order, update billing) completes the task.
Does not count:
- A chatbot that collects the question and dumps it on an agent. That is a routing step, not a resolution.
- A "deflection" where the customer gives up and churns. Abandonment is not resolution.
- A bot that loops the customer through dead-end menus until they rage-click "talk to a human."
The honest metric is resolution without a human, not contact avoided by any means. A bot that frustrates people into leaving is destroying deflection's whole point.
What is a good deflection rate
Benchmarks vary by product complexity, content quality, and channel, but here is the realistic shape of it:
| Setup | Typical deflection rate |
|---|---|
| Help center only | 20 to 30% |
| Knowledge base + AI agent | 40 to 60% |
| Narrow product, heavily templated | 60 to 70%+ |
If you are running a help center alone and seeing 20 to 30%, you are normal. The jump to 40 to 60% almost always comes from adding an AI agent that can read your documentation and resolve a conversation rather than just surface links. Rates north of 70% exist, but they usually mean a simple product or a deliberately narrow support scope.
Treat any vendor claiming 80%+ as the default with healthy skepticism. Ask how they count a resolution, and whether abandoned chats are quietly inflating the number.
How deflection turns into savings
Deflection is only worth talking about because of the economics. Every repetitive ticket an AI resolves is agent time you do not pay for.
Here is the trade. A human-touched ticket carries a loaded cost: agent salary, tooling, overhead, divided across the tickets they handle. For most teams that lands somewhere between $5 and $12 per ticket. An AI resolution, by contrast, is a flat metered cost. On SupportWire, each AI resolution costs $0.49, half of Intercom Fin's $0.99 per resolution.
So when the AI deflects a ticket that would have cost an agent $7 to handle, you spend $0.49 and keep the difference. Multiply that across hundreds or thousands of deflected tickets a month and the gap compounds quickly.
The mechanism that makes this work on SupportWire is auto-resolve: the AI reads from your knowledge sources, answers the customer, and closes the conversation when it is genuinely done, handing off cleanly when it is not. Pricing is per seat plus $0.49 per AI resolution, with a free plan and no sales call, so the deflection you earn shows up directly as margin rather than getting eaten by the tool that delivered it.
To see the number for your own volume, the ticket deflection calculator takes monthly tickets, your deflectable share, and your cost per ticket, then returns the net monthly and yearly savings after the $0.49-per-resolution cost.
Deflection vs avoidance
These two get used interchangeably, but they are different levers.
Deflection resolves a request through self-service or AI so it never reaches an agent. The question still gets asked, it just gets answered without a person.
Avoidance stops the question from being asked at all. You notice 400 tickets a month are "where is my tracking number," so you put the tracking link in the order-confirmation email and the tickets evaporate. No bot involved, the demand disappeared at the source.
The best support programs run both. Deflection handles the questions you cannot eliminate. Avoidance removes the ones you can. A rising deflection rate on a flat or shrinking ticket volume is the sign you are doing both well.
How to actually raise your deflection rate
Deflection is downstream of content and tooling. If your rate is stuck, the lever is almost always one of these:
- Fix the knowledge gaps. Pull your top 20 ticket reasons. Any that lack a clear, current article are pure deflection you are leaving on the table. An AI agent can only resolve what your documentation covers.
- Give the AI real resolution power, not just search. Surfacing links is not deflection. The agent needs to read the sources and answer the question in the conversation. That is the difference between a 25% help center and a 50% AI-backed one.
- Make handoff clean, not punishing. Customers tolerate a bot that resolves their issue and escalates instantly when it cannot. They do not tolerate a maze. Clean handoff protects trust, which protects your deflection numbers over time.
- Measure resolution, not contact avoided. Hold the line on what counts. If abandoned chats are inflating your rate, you are optimizing the wrong number.
Get those four right and the deflection rate takes care of itself. Then the savings are just arithmetic, the kind the calculator does for you in about thirty seconds.
Frequently asked questions
Ticket deflection is the share of inbound support requests resolved without a human agent. A customer finds the answer in a help article, a chatbot, or an AI agent and never opens a ticket for a person to handle. It is measured as deflected requests divided by total inbound requests, expressed as a percentage.
Ticket deflection rate = (requests resolved without an agent ÷ total inbound requests) × 100. If 900 of 2,000 monthly requests are resolved by self-service or AI, the deflection rate is 45%. To measure ROI, multiply the number of deflected tickets by your loaded cost per ticket, then subtract the cost of the automation that resolved them.
A help-center-only deflection rate usually lands around 20 to 30%. Teams that pair a solid knowledge base with an AI agent commonly reach 40 to 60% on repetitive, well-documented questions. Rates above 70% typically signal a narrow product or heavily templated support, and are the exception rather than the rule.
They overlap but are not identical. Deflection resolves a request through self-service or AI so it never reaches an agent. Avoidance is broader: it includes fixing the root cause so the question stops being asked at all, such as clearer UI copy or a better onboarding flow. Good support programs do both.
No. Deflection clears repetitive, low-value tickets like password resets and order-status questions so agents spend their time on the conversations that actually need a person. The goal is to redirect human effort to higher-value work, not to remove it.
Updated June 2026
