By the Onsites AI team · Last updated · 5-minute read
First contact resolution is the quiet metric behind every other queue number: the share of threads solved in a single exchange, with no "just following up," no reopen, no second ticket that re-explains everything. It is the metric that buys a small team its time back — every reopen is a thread the desk pays for twice, every half-answer a buyer who has to ask again — and it is the metric buyers feel most, because repeat contact is the definition of support that didn't work. This guide measures FCR honestly (it is subtler than "one reply"), names the three enemies that hold it down, and coaches it upward with the levers a free-tier desk actually has: the handbook, the reply discipline, and the authority to close things completely.
FCR's definition carries two traps that flatter every desk that falls into them. Trap one — counting first replies: a thread can be solved on second-and-third exchanges without a reopen (the buyer replies "that fixed it, also one more thing"); a desk counting replies overstates itself. The honest measure: solved in one exchange, confirmed by silence or a satisfied next message — count threads closed with no recontact inside seven days and no reopen. Trap two — counting closes: the reply that "resolves" by saying all the words while answering none of the question closes the ticket and reopens the buyer's day. Watch the paired metric: reopen rate over seven days is FCR's auditor — a desk whose FCR is 70% and whose reopens are 12% is lying to one of them. At small volumes, read FCR monthly as a trend, not as a verdict: 54% → 61% → 68% over a quarter is real movement; the decimal itself is not the point, the direction is.
The most common FCR killer is the reply that answers the words rather than the question. "My invoice #1041 looks wrong" answered with "invoices are generated on order confirmation" was the seller asking "why is €12,400 wrong?" — the half-answer closes 1 minute and reopens in 40. The fix is reading discipline: before replying, name the buyer's actual question in one clause ("they're asking why the price differs from quote #1042"); the reply-reading habits and the handbook's worked examples train it — the copilot drafts better when the desk's notes carry real questions, not surface ones. The half-answer's signature: replies that answer "how do I" with "what it is" — policy stated, next step missing. The complete answer states the thing, the next step and the date; the difference is one sentence and an entire relationship.
The buyer asks in one language (theirs); the desk answers in another (its own). "Has my order shipped?" answered with "the status was updated to fulfilled" — technically true, operationally a reopen, because "fulfilled" is the store's word and "shipped, arriving Thursday by carrier X, tracking here" is the buyer's. The fix is vocabulary translation, taught once: the handbook's saved replies in the buyer's words (the template discipline); the help center articles titled the way buyers ask, not the way systems name; the desk's shared glossary of its own jargon (statuses, stages, the difference between confirmed and paid). The wrong-vocabulary reply is a small-team disease because small teams talk to each other all day and forget buyers don't share the vocabulary — the audit that catches it: read last week's replies as a buyer would, and count the ones that needed translation.
The deepest FCR killer is the desk's own rules: the agent who could not refund, redate or requote without asking someone, so the "resolution" was a promise to find out — three touches for one problem, every time. The fix is written authority: the refund ladder empowered to the desk (the refund guide's four rungs); requotes from the rate card available to anyone answering (the quote-in-thread pattern); the delay note pre-approved by policy (the delay playbook). Authority is not a personality trait; it is written policy on what the desk can decide alone — and every decision taken away from the thread is a reopen scheduled in advance. The desk runs a monthly "what did I have to ask permission for?" list; each item named is authority granted or a process written, and FCR climbs the week the list shrinks.
FCR moves in response to named levers, reviewed as a ten-minute ritual (the KPI cadence). Read the reopens: every thread reopened inside a week gets one question — what would have closed it first time? The answers are the week's coaching: a macro missing (write it), a policy unknown (handbook it), a question misread (read-discipline note). Read the repeats: the topics that generate second and third contacts (shipping dates, deposit status, size questions) get handbook coverage with dated examples — the handbook's compounding is FCR's biggest lever, because a drafted-from-context reply closes what a template-from-memory half-closes. Watch the leverage line: the AI share of replies rises with review discipline intact (the quality lens), and every drafted-complete reply is an FCR the desk didn't have to find by hand. One fix per week, honestly measured: a small desk can move FCR fifteen points in a quarter, and every point is a quieter queue.
One desk's FCR quarter, numbers attached. Start: FCR 52%, reopens 14%, the reopens dominated by invoice queries and shipping dates. Week 2: the invoice-answer macro rewritten around the real question ("why is the amount different from the quote?" — with the quote-in-thread link as the answer); reopens on invoices drop from six a week to one. Week 5: shipping replies re-shaped to lead with the date and the tracking, authority granted to send delay notes on policy; shipping reopens halve. Week 9: the handbook's top-ten questions carry worked examples; the copilot's drafts start closing more threads first touch, and the AI share of replies — at credit prices that keep the line in single digits (published) — reaches 60%. Quarter end: FCR 69%, reopens 5%, and the team's daily queue visibly quieter — the same 600 monthly threads now cost four fewer hours weekly of rework. The quarter cost: twelve ten-minute rituals and a handful of written policies. That is FCR's whole business case for a small desk: buy the time back once, keep it.
What is first contact resolution?
The share of threads solved in a single exchange — no follow-up contact, no reopen. Measured honestly: solved-in-one-touch confirmed by no recontact within seven days, with the seven-day reopen rate as its auditor (a 70% FCR with 12% reopens is lying to one of them).
How do you measure FCR on a small desk?
Monthly trend, not decimal-per-agent. Count threads closed with one exchange and no recontact in a week, divide by resolved threads. Read reopens at the same time — every reopen is a first answer that half-closed, and it is where the coaching lives.
What keeps FCR low?
Three enemies: half-answers (replying to the words, not the question), wrong-vocabulary replies (desk terms instead of buyer terms), and missing authority (the agent who can't refund, requote or redate alone, so every resolution is a promise to find out). Each is fixable by written discipline, not heroics.
How do you coach FCR upward?
Weekly ten-minute ritual: read the week's reopens and ask what would have closed them first time; write the missing macros, handbook the unknown policies, grant the missing authority. One named fix per week compounds — small desks routinely move FCR 10–15 points a quarter.
Does AI help first contact resolution?
Directly: drafts grounded in the handbook and the thread's context (orders, quotes, accounts) close questions completely on first reply, at credit prices that keep the line in single-digit dollars. Watch the AI share of replies rise alongside review discipline — drafted-complete is FCR the desk didn't have to find by hand.