Deflection Rate: How to Measure It

Deflection rate is the metric form of ticket deflection: a single percentage that answers "what share of incoming support did we resolve without a human?" Where ticket deflection names the outcome, deflection rate puts a number on how often that outcome happens across all your conversations in a period.

At its simplest the formula is the count of conversations resolved by self-service or automation, divided by the total count of conversations that started. A low rate on one platform and a high rate on another may describe identical support quality. The gap is usually in what each side chose to put in the numerator and the denominator.

The denominator problem

Deflection rate is unusually easy to move without improving anything, and almost always by editing the denominator. Exclude "browsing" sessions, or sessions under a few seconds, or conversations that never reached the bot, and the rate climbs while the customer experience is unchanged. The opposite games it too: counting every page view as a deflected ticket inflates the number past the point of meaning. Because the formula is so sensitive to these choices, a deflection rate is only comparable to itself, measured the same way over time, never across two vendors' differently drawn definitions.

Rate versus resolution

A high deflection rate is not automatically good. A conversation can be "deflected" because the customer gave up, not because they were helped — a false deflection that hides a failure and often produces a second, angrier contact later. This is why deflection rate should always be read next to a satisfaction or resolution signal. A rate that rises while CSAT falls is not deflection; it is abandonment wearing deflection's clothes.

Measuring your own rate

There is no meaningful benchmark deflection rate to aim for; the useful figure is your own, tracked consistently. Fix your definition of a countable conversation and a countable deflection, apply it unchanged, and watch the trend and its by-topic breakdown rather than a single headline. Improvement should come from answering more questions correctly, not from redrawing the denominator.

Deflection rate in Evoriqa

Evoriqa reports AI deflection alongside resolution rate and CSAT on one analytics dashboard, measured on your real conversations rather than an assumed baseline. Because its agents answer from your knowledge and hand off with context when they cannot, a deflected conversation reflects a question actually resolved, and the CSAT shown next to it is the check against a rate that looks good for the wrong reason. Knowledge-gap detection surfaces the questions dragging your rate down, so you raise it by adding answers, not by adjusting the maths.