Here is how I build a ticket deflection strategy for B2B support. A ticket deflection strategy is a plan for answering repeatable questions through self-service at the moment customers look for help, while keeping a person one click away for account-specific, urgent or risky issues. Measure it honestly: deflection rate = confirmed deflections ÷ (confirmed deflections + tickets created) × 100. Moving from 20% to 30% on 6,000 help requests a month means 600 fewer tickets.
Price what those tickets cost with the free cost per ticket calculator. In Part 5 we reduced escalations by giving tier 1 what it needs. This part moves one step earlier, to the questions that never needed an agent.
What counts as a ticket deflection?
A deflection happens when a customer who meant to contact support finds the answer and does not submit a ticket. The key words are "meant to." A help center page view is not a deflection, because most readers were never going to write in.
Count only customers who start a contact flow, such as opening the help widget or contact form, see a suggested answer and leave. A confirmed deflection is one where the same customer does not open a ticket on that topic within 7 days.
How do you calculate deflection rate?
Deflection rate (%) = Confirmed deflections ÷ (Confirmed deflections + Tickets created) × 100
Confirmed deflections = customers who started a contact flow, saw an answer,
left, and did not open a ticket on that topic within 7 days
Tickets created = tickets submitted through the same flows in the period
Some help desk vendors use a broader "self-service score": help center users divided by users who submitted tickets. Zendesk describes it that way. It is useful for trend, but it counts every reader, so I do not report it as deflection.
Free download
Get the Erlang calculator (call center staffing) spreadsheet
The Erlang calculator (call center staffing) as an Excel file with the formulas set up, ready for your own numbers. Works in Google Sheets too.
The download button appears here right away; nothing is sent to your inbox. Your email also joins Clarivoxx Insights (a weekly article and a monthly newsletter). Unsubscribe in one click. Privacy
What does deflection look like with real numbers?
Say a team sees 6,000 help requests a month. Today 1,500 customers leave after viewing a suggested answer, but 300 of them come back with a ticket within 7 days.
| Step | Calculation | Result |
|---|---|---|
| Confirmed deflections | 1,500 − 300 | 1,200 |
| Tickets created | 6,000 − 1,200 | 4,800 |
| Deflection rate | 1,200 ÷ 6,000 | 20% |
| At a 30% target | 1,800 deflections | 4,200 tickets |
That is 600 fewer tickets a month. At about $16 per ticket, it saves $9,600 a month. Run your own cost in the cost per ticket calculator. The bigger gain is time for the complex queue, the same capacity you need to reduce ticket backlog.
What 10 points of deflection are worth
Deflection rate = confirmed deflections ÷ (confirmed deflections + tickets created) × 100. Here 1,200 ÷ 6,000 = 20% today and 1,800 ÷ 6,000 = 30% at target, so tickets fall from 4,800 to 4,200, and 600 × $16 = $9,600 a month.
Why is B2B deflection different?
Customers are fewer and worth more, many questions depend on account setup, and large accounts expect a person when something breaks. Customers do try self-service first: in research published in Harvard Business Review in 2017, 81% of customers across industries tried to solve the issue themselves before reaching a live representative. The job is to make that first attempt work, not to block the second.
What should you deflect, and what should reach a person?
Group 90 days of tickets by category and ask three questions:
- Is the answer the same for everyone? How-to steps usually are. A failed data sync usually is not.
- Does it need account data? Then self-service can only collect context.
- What if the answer is wrong? Security, data loss and outages should reach a person fast.
Which ticket categories self-service can realistically absorb
Focus content on the top two categories first; bug reports belong with a person.
You end up with three groups: deflect repeatable questions, assist where self-service guides and then hands off with context, and route to a person anything urgent, risky or commercial, such as renewals.
How do you build a ticket deflection strategy?
- Start from ticket drivers. Write or rewrite articles for the top 20 contact reasons first.
- Put answers where the question is asked. Suggest articles in the contact form as the customer types, and add help links on the screens that create the most tickets.
- Fix search. Review zero-result searches every week.
- Use AI answers with limits. Use only your published articles as the source, show which one, and hand off with the conversation attached.
- Publish a status page so incidents do not turn into hundreds of "is it down?" tickets.
- Keep the contact button visible. Pass the article viewed into the ticket so nobody repeats themselves.
- Give every article an owner and review the top ones monthly.
Deflect, assist or route to a person
Deflect with self-service
How-to and navigation questions
Password resets
Invoice downloads
Plan limits and features
Assist, then hand off with context
Integration setup
Data imports
Billing changes
User permission questions
Route to a person quickly
Suspected bugs and outages
Security or data-loss concerns
Renewal and contract questions
Issues from strategic accounts
The runbooks tier 2 wrote to reduce escalations are good candidates for customer-facing versions once the steps are proven.
How do you know deflection is really working?
A higher deflection rate is only good news if customers really get answers. Track alongside it:
- False deflection rate: returns within 7 days ÷ customers who left after an answer. Here, 300 ÷ 1,500 = 20%.
- Tickets per 100 active accounts, so growth does not hide progress. The contact rate calculator does this split.
- Effort on articles, using the customer effort score calculator and the method in how I calculate customer effort score.
What I've seen in practice
In travel tech, I learned that refund questions sit right on the deflect line. "How long does a refund take?" is the same answer for everyone and belongs in a short article. "Where is my refund for booking X?" needs a person. When we split the two in the contact form, the general question mostly stopped reaching agents, and the specific ones arrived with the booking details already attached.
On one team I worked with, a new help widget hid the contact button two clicks deep. The deflection rate looked great for a month. Then partners started emailing their account managers directly, and those threads were harder to track than the tickets ever were.
A deflection dashboard you can copy
I review these five rows together every month, using the numbers from the worked example.
| Metric | Target | This month | Last month | Change |
|---|---|---|---|---|
| Deflection rate | 30% | 20% | 18% | +2 pts |
| Tickets created | 4,200 | 4,800 | 4,756 | +44 |
| False deflection rate | 15% or less | 20% | 22% | −2 pts |
| Tickets per 100 accounts | 168 | 192 | 195 | −3 |
| Ticket cost | $67,200 | $76,800 | $76,096 | +$704 |
Total tickets went up, but tickets per 100 accounts went down, so growth is hiding real progress. Read deflection and false deflection together: if both rise, customers are leaving without answers and will come back another way.
What to do this week
List your top 20 contact reasons and mark each one deflect, assist or route. Pick the three easiest "deflect" items and check whether a short, current article exists for each.
Frequently asked questions
Is a help center page view a deflection?
No. Most readers were never going to contact support. Count only customers who started a contact flow, saw an answer and did not open a ticket.
What is a false deflection?
A customer who left after seeing an answer but came back with a ticket on the same topic within 7 days. A high false deflection rate means the articles do not answer the question.
Should AI chatbots count toward deflection?
Yes, under the same rule: the conversation counts only if the customer does not open a ticket on that topic within 7 days.
Should every B2B question be deflectable?
No. Outages, security concerns, renewals and issues from strategic accounts should reach a person quickly.
Sources
- Kick-Ass Customer Service, Harvard Business Review
- Ticket deflection: enhance your self-service with AI, Zendesk
Run your own numbers
Free, no sign-up, with the formula and an Excel version.
Read next
How to Build a Support Team Scorecard: The Metrics That Matter
A support team scorecard with 8 metrics to review weekly, each with a formula, a target and an owner, plus a worked week and a 30-minute agenda.
Read the article
Go further with the ebook
The Professional Customer Support Playbook
Better replies, fewer escalations, a clear path from agent to leader.
$19.99 · 103 pages · View the book