Measuring Live Chat Deflection Rate

What deflection actually measures

Deflection rate is the share of questions that live chat resolves before they become something heavier — a phone call, an email thread, a formal ticket, or a lost customer. It is a proxy for how much cheaper, faster work chat is absorbing. Like any proxy, it is useful right up until you start gaming it, so define it carefully before you celebrate it.

Define it honestly

The honest version counts a chat as deflected only if the customer's issue was actually resolved in chat and did not resurface through another channel. The dishonest version counts every chat that did not immediately become a ticket — which quietly includes the frustrated customer who gave up and churned. If your deflection rate and your re-contact rate both look great, trust it; if deflection rises while re-contacts also rise, you are not deflecting, you are deferring.

What pushes it up the right way

• Answers that actually resolve: a chat that ends with a working solution, not a “that should do it.”
• Good self-serve behind it: agents who link the right help article turn one deflection into many.
• Fast enough response: people escalate to phone or email when chat is too slow; speed keeps them in the cheaper channel.

A high number is only good if satisfaction holds

Deflection is a cost metric, and cost metrics are dangerous alone because the cheapest possible support is no support. A rising deflection rate paired with falling CSAT means you are pushing people away, not helping them. The search-side counterpart is worth reading alongside it, since help center search analytics show what people looked for and did not find. Always read deflection next to a satisfaction signal; the pair tells the truth that either one alone will hide.

How MyLiveChat helps you measure it

MyLiveChat's transcripts, tags, and reporting let you mark how chats resolved and spot the ones that came back, and its post-chat ratings give you the satisfaction signal to read alongside. That combination is what lets you claim a deflection number you can actually defend — resolved and satisfied, not just “didn't open a ticket.”

Write the definition down before you report a number

Deflection is unusually easy to measure badly, because every term in it is a choice. Two teams reporting the same figure can be measuring entirely different things, so the definition belongs in writing next to the number.

• What counts as deflected? A chat that closed without a ticket, or one that closed without any further contact from that customer for a set window? The second is far more honest and much harder to game.
• Over what window? Seven days is a reasonable default. Without a window, a customer who emailed the next morning still counts as deflected.
• Which chats are eligible? Sales enquiries were never going to become support tickets. Including them inflates the number for free.
• Do abandoned chats count? A visitor who left mid-conversation did not have their issue resolved. Counting them as deflected is the single most common way the metric lies.

The ways this metric goes wrong

Deflection has a structural flaw: the easiest way to improve it is to make other channels harder to reach. Hide the support email, remove the phone number, and deflection rises while customer experience falls. Any team measured on this number should also be measured on satisfaction and repeat contact, or the incentive quietly points the wrong way.

The second failure is counting chats that were never tickets in the first place. If chat attracts a new population of quick questions people would previously have shrugged off, those conversations are additional volume, not deflected volume. They may still be worth having — they often prevent an abandoned purchase — but they belong in a different number.

The third is the unresolved close. A chat that ended because the visitor gave up looks identical in the data to one that ended because the problem was solved, unless you deliberately separate them.

A calculation you can defend

1. Start with support-intent chats in the period, excluding sales and excluding chats that ended without an agent reply.
2. Remove any chat where the agent explicitly created a ticket or escalated.
3. Remove any chat where the same customer contacted any channel again about the same issue within seven days.
4. What remains is deflected. Divide by the eligible total from step one.

Step three is the one teams skip because it needs identity matching across channels. It is also the step that turns the metric from a flattering number into a useful one. If you cannot do it yet, report the figure explicitly as chats closed without a ticket rather than calling it deflection.

What actually raises it honestly

Resolution authority is the biggest lever. Agents who can issue a refund, resend an order or reset something finish the conversation; agents who can only take details are a routing step with extra latency. Second is a good knowledge base the agents themselves use, so the answer is accurate and consistent. Third is fixing the source: the recurring question that keeps arriving is usually a page that should have answered it.

What to measure alongside it

Never report deflection alone. Pair it with satisfaction on the same chats, with repeat contact within the window, and with the ticket volume it is supposed to be reducing. If deflection climbs while tickets stay flat, the chats were probably additional demand rather than diverted demand — which is worth knowing before anyone claims a saving.