The 30-second answer

If 100 customers contact you about something and 70 of them don't come back about the same thing, your FCR is 70%. That's the calculation. The interesting question is what counts as "don't come back" — and most teams pick the easy answer.

The easy way most teams measure it

Two common shortcuts, both inflate the number:

  • Agent self-report. At the end of the call, the agent ticks a box: "resolved on first contact?" Yes. The number is high because the agent's bonus depends on it.
  • Single-channel only. The chat report says the chat ended; the email a day later about the same issue lands in a different system and doesn't count. FCR looks great in chat, the real customer experience is worse.

Both methods produce numbers that the front-line team likes and the customer doesn't recognise.

The honest way

An issue counts as resolved on first contact only if both of these are true:

  • The customer didn't come back about the same issue through any channel within a defined window (24, 48 or 72 hours).
  • If a survey went out, the customer marked it solved.

That's the same construction as resolution rate, applied per-contact instead of per-conversation. The two metrics are close cousins — FCR is the version a traditional contact centre uses; resolution rate is the version that comes up more in AI-agent vendor pitches. Both try to answer the same question: did the customer get what they came for?

Why FCR overstates more than it understates

Most measurement errors push FCR up. The agent forgot to tag the issue as repeat. The email three days later got routed to a different agent and recorded as new work. The customer gave up and switched providers — which doesn't show in your data at all. None of those reduce reported FCR; all of them mean the real number is lower than what the dashboard says.

When a vendor or a consultant quotes you an FCR figure, ask: was it self-reported by the agent, or measured by recontact across channels? The difference is usually 10-20 percentage points.

How AI affects FCR

An AI agent that handles 60% of contacts changes which conversations a human even sees. The human agents end up working on the harder remaining 40% — where FCR is naturally lower because the problem was already complex enough to escalate.

So if you compare "AI FCR" to "human FCR" on the same contact pool, the AI looks better. On a fair comparison (matched complexity), it's usually closer or actually worse on the edge cases. The honest framing isn't "AI improves FCR"; it's "AI handles the easy contacts at very high FCR and leaves the hard contacts in the human queue."

What good looks like

  • 75-85% (cross-channel, honestly measured) — a healthy contact centre.
  • 85-95% — either exceptionally good operations, or someone is measuring loosely. Stress-test the methodology.
  • Below 70% — there's structural friction (broken knowledge base, fragmented systems, agent training gap) worth fixing before any AI investment.