What is deflection rate?
Deflection rate measures the percentage of customer enquiries an AI assistant can resolve without human intervention. It deflects routine queries away from your teams, so they have more time to work on higher-value tasks, which can significantly increase efficiency, boost team morale and improve customer satisfaction.
Increasing your deflection rate can help improve other customer service metrics, like Net Promoter Score (NPS), goal completion rate and average handling time, as well as reduce your cost per call and escalation rate, since customers are getting immediate, accurate, helpful responses to their queries from an AI-powered assistant 24 hours a day in 130+ languages with real time engagement that’s both prompt and convenient.
How do you calculate your deflection rate?
To calculate the deflection rate, divide the number of enquiries resolved through self-service tools like AI assistants (R) by the total number of enquiries received (E), and multiply by 100 to get a percentage.
Deflection rate formula
R ÷ E x 100 = deflection rate %
For example, 98 queries out of 100 in total resolved entirely using a next-gen AI assistant:
98 ÷ 100 x 100 = 98% deflection rate
Calculation frequency
You can calculate your deflection rate monthly, weekly or even daily. It depends on your industry and approach, if your business is seasonal or cyclical, if you have product or service releases that invite a lot of customer questions, or you’re making changes to your organisation ― you decide. The most important thing is to calculate your deflection rate consistently over time using the time span that gives you the most valuable insights into the effectiveness of using self-service options and how this impacts your business.
By tracking and monitoring the deflection rate regularly, your teams can gauge the performance of your AI assistant accurately and so identify areas where you can improve it. Comprehensive AI-driven data can transform your customer insights and help make sure you’re providing the best experience possible.
Why does the deflection rate matter?
Deflection rate directly impacts your operational efficiency, how you allocate your resources, and your customer satisfaction score. That’s because using AI automation leads to a faster first response time, which is good news for customers, but deflecting routine enquiries is also good for your human agents.
By improving the deflection rate, you open up many opportunities for your support teams to enjoy non-stressful interactions with customers, and when an AI assistant is handling the monotonous, repetitive queries, your teams are more available to focus on the sensitive, challenging or unusual enquiries that need their skills and experience to resolve them before the customer becomes unhappy or irate.
The more your AI assistant learns about your organisation (directly through your customers via the topics they most want to talk about), the more it will be able to do for you. Not just sending information, but through integration with all your go-to business systems, like HubSpot and Jira, it can carry out tasks and transactions for you too, dealing with enquiries end-to-end, further increasing your deflection rate.
Reducing customer service costs
With the average cost per call for contact centres in the UK sitting at £6.26, you’ll immediately reduce your operational costs when you increase your deflection rate, since the cost of an AI-powered response is mere pence (you can estimate your saving upfront using our handy calculator).
London’s Barking & Dagenham Council saved £48,000 in their first six months after introducing an AI assistant to handle missed bin enquiries, and leisure providers Mytime Active saved the cost of having to hire two extra agents to deal with routine enquiries.
What is a good deflection rate benchmark?
In the past, a good deflection rate could have been as low as 20% and peak at 50%, but a new generation of AI assistants with deeply complex features like natural language processing (NLP) and large language model (LLM) capability has changed this for good and our AI assistants average a 96% success rate.
A 96% success rate means the AI assistant has understood nearly all of the requests from your customers and been able to resolve their issues without any input from your contact team, simply deflecting them away. With more training, AI assistants can get even higher rates, like Mytime Active currently at 97% and both Legal & General Insurance and Barking & Dagenham Council at 98%.
The 3 do’s and don’ts to consider
To find a level of success as high as 96%, there are three important things to note in the age of AI automation:
1) DON’T overload your AI assistant
You can launch an advanced AI assistant in less than ten minutes and it can go live with as little as three main flows (flows are what we call a question and answer set). You don’t need to train it on everything you think your customers will want to know because as soon as your AI assistant is live and chatting with real customers, it immediately starts to learn this and deflect those queries away. Simply set it live and let it learn.
2) DON’T aim for a 100% success rate
There will always be questions asked of your AI assistant that aren’t relevant to what you do or are simply better off handled by one of your team in person over live chat or the phone, so 100% isn’t a necessary or realistic goal for your success rate. You’ll also need to account for these non-queries in your deflection rate, since they’re not enquiries you’ll ever want to train your AI assistant to answer or deflect.
3) DO keep scaling your AI assistant (and never stop)
Your deflection rate should continue to go up over time and you can help this along by using integration to make your AI assistant smarter. Connect up your AI assistant with all your favourite tools, like Jira, HubSpot or Zapier, and it’ll be able to deflect more queries for you by completing tasks and transactions on your behalf, becoming more and more valuable to your teams every day.
