How to calculate ticket deflection rate from documentation

If you add a chatbot to your documentation, you may want to know whether it reduces support tickets.

A simple way to measure that is to compare support demand before and after the chatbot is introduced.

Start with the overall ticket rate. Then break it down by topic to see where the chatbot and documentation are helping, and where readers still need support.

Calculate the ticket rate before and after

Use:

Ticket rate =
support tickets ÷ documentation and chatbot interactions × 100

For example, before adding a chatbot:

Documentation interactions: 20,000
Support tickets: 1,000
 
Ticket rate = 1,000 ÷ 20,000 × 100 = 5%

After adding the chatbot:

Documentation and chatbot interactions: 25,000
Support tickets: 875
 
Ticket rate = 875 ÷ 25,000 × 100 = 3.5%

The ticket rate fell from 5% to 3.5%.

You can also calculate the relative reduction:

Ticket-rate reduction =
(before rate − after rate) ÷ before rate × 100
 
Ticket-rate reduction =
(5% − 3.5%) ÷ 5% × 100 = 30%

In this example, support tickets fell by 30% relative to documentation and chatbot usage.

Keep the underlying numbers with the percentage so the result is easy to understand.

Look at the overall number first

An overall ticket rate gives you a simple top-level metric.

This helps answer the broad question:

Are fewer documentation users reaching support after we added the chatbot?

It is especially useful when documentation traffic changes. Raw ticket volume alone can be misleading if your audience is also growing.

Break the result down by topic

The overall number tells you the direction. Topic-level numbers tell you what is causing it.

For example:

Here, authentication improved significantly, integrations barely changed, and billing got worse.

That gives you somewhere useful to investigate.

You might discover that:

• authentication questions are now answered well by the docs and chatbot
• integration documentation is still missing important steps
• billing questions require account-specific support
• a product change created new confusion

You do not need to measure every topic from day one. Start with the areas that generate the most support demand.

Make the before-and-after comparison fair

Try to compare periods that are reasonably similar.

Check whether something else changed at the same time, such as:

• a major product release
• an outage
• a migration
• new ticket categories
• a different customer mix
• a new onboarding flow
• a large change in traffic

If you launched a chatbot during the same week as a major product migration, for example, the change in ticket volume probably cannot be attributed to the chatbot alone.

You can still report the numbers. Just keep the context with them.

After launch, look at conversations that still lead to support

The before-and-after comparison tells you whether support demand changed.

Once the chatbot is live, another useful question is:

Which chatbot conversations are still followed by support tickets?

If you can connect chatbot activity with your support system, calculate:

Post-chat ticket rate =
chatbot conversations followed by a matching support ticket
÷ chatbot conversations
× 100

For example:

Authentication chatbot conversations: 800
Conversations followed by an authentication ticket: 72
 
Post-chat ticket rate = 72 ÷ 800 × 100 = 9%

That means 9% of the measured authentication conversations were followed by a related support ticket.

Do not assume the other 91% were all successfully deflected. Some readers may have solved the problem, while others may have left without completing the task.

The useful part is the 9% you can investigate.

Review why readers still open tickets

Look at the conversations that were followed by support and ask what was missing.

Common reasons include:

• the relevant page was hard to find
• the documentation did not cover the task
• an important prerequisite was missing
• an example was outdated
• the chatbot used the wrong source
• the answer did not give the reader enough information to continue
• the reader needed account-specific help
• the problem was in the product rather than the documentation

Then improve the relevant source and measure the same topic again.

For example:

Before documentation fix
 
Integration conversations: 500
Matching support tickets: 70
Post-chat ticket rate: 14%

After fixing a missing setup step:

After documentation fix
 
Integration conversations: 550
Matching support tickets: 44
Post-chat ticket rate: 8%

Now you have a simple improvement loop:

1. measure which topics still reach support
2. review the conversations and documentation
3. fix the underlying problem
4. measure again

Use Biel.ai activity with your support data

Biel.ai Analytics can help you understand what readers are asking your chatbot, including sessions, messages, common questions, Content Gaps, and conversation origin URLs.

See the Analytics documentation for current metric definitions and availability.

Your support system provides the ticket data.

Together, the two can help you see whether support demand changes after adding a chatbot and which reader questions still need work.

Frequently asked questions

How do you calculate ticket deflection from documentation?

Compare support tickets with documentation and chatbot usage before and after introducing the self-service experience:

Ticket rate =
support tickets ÷ documentation and chatbot interactions × 100

A lower rate after launch means fewer support tickets relative to self-service usage.

Should we measure all support tickets or individual topics?

Both can be useful. Start with an overall rate for the top-level trend, then break it down by topic to see where support demand is actually changing.

Is every chatbot conversation without a ticket a deflected ticket?

No. You only know that no matching support ticket followed the conversation during the period you measured.

Should we use ticket count or ticket rate?

Use the rate when documentation or chatbot usage changes between periods. Ticket counts are still useful, but the rate gives you context about how much self-service activity occurred.

What should we measure after the chatbot is live?

Look at which chatbot conversations are still followed by support tickets. Those conversations can show you where the documentation, chatbot answers, or product experience still need improvement.

Start with the overall rate

Measure documentation usage and support tickets before adding the chatbot.

Measure them again after launch.

Then break the result down by your largest support topics and inspect the chatbot conversations that still lead to tickets.

That gives you a simple way to track whether support demand is falling and a practical way to decide what to improve next.

If you want to add a chatbot to your documentation and start collecting usage data, create a Biel.ai project and follow the Quickstart to connect your documentation.