What it means
Customers may find answers in a help centre, chatbot or account portal instead of contacting an agent, and a business may call that ticket deflection. The rate tries to describe how much demand these paths keep out of the support queue.
Defining demand is difficult, because a page view does not prove the visitor had a problem and leaving without creating a ticket does not prove success. One model divides estimated successful self-service interactions by those interactions plus tickets created.
If 3,000 genuinely confirmed self-service resolutions and 7,000 comparable tickets are recorded, the illustrative rate is 30%, and the word "confirmed" matters. Another system may use an estimated deflection count based on article views and ticket creation, which is a proxy with assumptions, so label the method whenever numbers are presented.
Zendesk distinguishes deflection from resolution: keeping a case out of a human queue does not establish that the person got the answer, and it recommends looking at repeat contact and outcomes. A simple feedback button gives some signal, but only a fraction of visitors may respond, so combine it with other evidence.
Follow-up windows matter too, since someone who reads an article and opens a ticket an hour later was probably not helped by it. A chatbot may answer a routine delivery question, but it can also give the wrong answer or block escalation.
Audit answer quality and make human help reachable, because raising deflection by hiding contact routes is not a service improvement. The target should also fit the issue type: password reset or order-status queries may suit self-service, while sensitive complaints may need a person.
Good knowledge content is accurate, findable and current. If an article describes an old product, customers may avoid filing a ticket yet act on bad information, so owners should review content when policies change.
Segment the measure by journey and keep help-seeking interactions in the denominator rather than all site visits. Track support cost and volume context carefully.
A smaller human queue can reduce work, but content production, software, monitoring and difficult escalations still cost money, and a rate may rise simply because ticket creation fell in a quiet season. When comparing periods, keep the same tagging, window and channel scope, and pair deflection with confirmed resolution and customer experience so that fewer tickets do not masquerade as better support.
In practice
Real-world examples.
Example
A support team confirms 3,000 self-service resolutions and handles 7,000 comparable tickets. Under its stated model, deflection is 30%. The team publishes the method, the window and the counts alongside the percentage.
Example
A visitor reads a help article and contacts support shortly afterward. The team does not count that article view as a successful resolution, because the follow-up ticket shows the article did not answer the question. The deduplication rule links the two events to the same customer.
Example
A chatbot sends people through a long loop without solving their problem. Ticket volume falls but customer effort rises, and repeat contacts increase the following week. The team treats the lower ticket count as a warning, not a success, and restores a visible route to a human agent.
Formula
Calculation
Illustrative confirmed self-service share = confirmed self-service resolutions / (confirmed self-service resolutions + comparable human tickets) x 100. For 3,000 / 10,000, the result is 30%; estimated deflection uses other assumptions.
Suppose the next quarter records 4,500 confirmed self-service resolutions and 6,000 comparable tickets. The total is 4,500 + 6,000 = 10,500, so the share is 4,500 / 10,500 x 100 = 42.9%. If the cost of handling a ticket is $8, the 4,500 resolutions avoided roughly 4,500 x $8 = $36,000 of agent handling, before subtracting the cost of content, software and monitoring.Case study
Seen in the real world.
This entirely fictional case follows Elm Parcels, an invented delivery business. It improved an order tracker and asked users whether their task was complete. The team monitored repeat contacts and unresolved issues alongside ticket volume. No real reduction in workload is claimed.
The support lead kept the contact button visible on every help page, even though hiding it would have lifted the dashboard rate. She reported the confirmed share, the repeat-contact rate and the customer feedback score on the same page each month. The illustrative lesson was that a falling ticket count only counted as progress when repeat contacts did not rise with it.
Watch out
Common mistakes.
- Counting every help-page view as a ticket successfully avoided.
- Hiding contact options to make the dashboard rate rise.
- Treating deflection as proof of resolution without outcome evidence.
Questions
People also ask.
Is deflection the same as solving a case?
No. It indicates a human ticket was avoided or estimated avoided; resolution needs separate evidence.
What should be counted?
Clearly defined help-seeking interactions under one measurement method and follow-up window.
Can a high rate be bad?
Yes, if people cannot reach help or their issues remain unresolved.
From the founder's library
Take it further with the book.
Build your financial confidence beyond this definition. Shihan's full-length guide, Accounting Fundamentals, takes the same plain-English approach and turns it into a complete, practical playbook for non-finance managers, business owners and students - with chapter-end quiz answers and presentation slides included.
25% off with code MMHQ25, applied at checkout. Priced in USD - checkout may show the equivalent in your local currency.
View the book and save 25%