What good looks like: support benchmarks by team size and industry
National satisfaction has been flat for years, phone beats chat, and a 'good' NPS in SaaS would sink a bank — why the only honest benchmark is your own segment.
Someone always asks it in the Monday review. The dashboard is up, the CSAT number sits there in a big font, and a manager two levels removed leans in: "Is that good?" The honest answer — "compared to what?" — never lands well. So you reach for a benchmark, and that is where the trouble starts.
Benchmarks feel like solid ground. They are mostly quicksand. A "good" number on the phone is a mediocre one in chat; a world-class score in retail would get you a stern meeting in financial services. Before you paste a figure into a slide, it helps to know what the ground actually looks like.
The national baseline barely moves
Start wide. Across the US economy, the American Customer Satisfaction Index puts national CSAT at 76.9 out of 100 — and it has been essentially flat, no materially higher than it was in 2017. That is the backdrop every team measures itself against: not a rising tide, but a plateau. If your satisfaction number is holding steady year over year, you are not stagnating. You are matching the entire country.
Channel changes the picture immediately. Phone support, measured on its own, tends to run hotter — industry-average call-center CSAT sits around 78%, because a resolved voice call carries a warmth a closed chat rarely matches. Compare a phone team to an email team on raw CSAT and you are not measuring quality. You are measuring the medium.
The operational metrics have real numbers
Satisfaction is the outcome; the operational metrics are the levers. Here the call-center world has the tightest benchmarks, because it has been measuring itself the longest.
First-contact resolution is the one worth anchoring on. The cross-industry average lands here:
Anything in the 70–79% band counts as good; 80% and up is world-class. FCR tracks satisfaction more reliably than almost any other operational number, which is why it belongs on the wall and average handle time does not.
Speed matters too, but less than people think. The benchmark average speed of answer is: