Last updated: September 12, 2026
What first call resolution measures
First call resolution is the share of calls where the caller's problem got handled without a second contact. The formula divides calls resolved on the first contact by calls handled in the same window. Both halves of that fraction are arguable, which is why two centers can run the same floor and report numbers twenty points apart.
Nobody agrees on resolved. One center counts a call resolved the moment an agent picks a disposition code at wrap up. The next one waits seven days and counts it resolved only if that phone number never came back. Same scripts, same people, same month, and the seven day version nearly always lands lower, because an agent at wrap up is predicting the future while the seven day version is checking.
The rate still gets quoted in board decks and client quality reviews as though the definition traveled with it.
The formula, and what goes in the denominator
First call resolution rate (FCR) = calls resolved on the first contact, divided by calls handled in the same window, times 100.
Everyone argues about the numerator. The denominator is what quietly decides the answer.
| Denominator | What it includes | What it does to the rate |
|---|---|---|
| All inbound calls handled | Wrong numbers, hangups, calls nobody could have resolved | The honest baseline, and the lowest of the four |
| Calls eligible for resolution | Only calls your written rule says could end on the phone | Raises the rate, and holds up only if that rule exists in writing |
| Unique issues rather than calls | Three calls about one problem collapse into one row | Closest to the customer's experience, and the most work to build |
| Contacts across every channel | The chat and the email that followed the call | Lowest number, and the only one that catches a caller who gave up |
A mixed channel queue should run the contact version of the same formula, and some teams publish the complement instead, one minus the repeat contact rate, where the arithmetic is identical and the emphasis sits on failures.
The word resolved needs translating before some floors can calculate anything. An operator taking a message for an HVAC contractor cannot fix the caller's problem, so resolution there means the message reached the on call tech with every field the account requires, which is why the numbers answering services run on treat accuracy as the outcome. A law firm intake desk resolves a call by booking the consult. Write that down before the arithmetic.
Three ways to decide a call was resolved
| Method | What it costs | What it gets wrong |
|---|---|---|
| Agent disposition at wrap up | Nothing, the codes sit in your platform already | Reads high, since the agent is guessing with wrap time running |
| Post call survey | Survey tooling, plus every caller who ignores one | Captures people at the two emotional ends of the experience |
| Repeat contact from platform data | Engineering work to match a caller against the previous week | Misses the caller who gave up, miscounts a second call about something new |
Asking the customer looks like the clean answer, and we half believe it. A survey asks the question directly, which nothing else there does, then hands you the view of the small group who answer surveys. Treat it as a check on the repeat contact number rather than as the number you publish.
Repeat contact measurement is the one worth the engineering. Matching on phone number alone scores a caller with two separate problems as a failure and a caller borrowing a friend's phone as a success, so build the match on the customer record.
The same month, measured three ways
Say a floor handles 10,000 calls in March. Agents code 8,900 of them resolved at wrap up, so the disposition version reports 89%. Match those same numbers against the previous seven days and 3,000 calls were followed by another call from that caller about the same thing, which puts the rate at 70%. Stretch the window to 30 days, another 400 callers turn up, and it lands at 66%.
Nothing about the floor changed between those three numbers, and nobody cheated. Pick the one you can still explain in a year, then print the window and the denominator rule beside the rate everywhere it appears, client reports included.
What inflates it
Disposition codes do the heaviest lifting, and the mechanism is ordinary. An agent who just spent nine minutes on a hard call has forty seconds of wrap time and a dropdown where resolved sits near the top. Eligibility rules take the next chunk, since a floor measuring against eligible calls can report 88% while resolving 62% of what rang in.
After that it is the contacts you cannot see. Somebody coming back through chat, or through the main line rather than the direct number, never shows up in a call level repeat match. The caller who quietly gives up is worse, since every method above except the survey scores that one as a win.
First call resolution against handle time
These two move against each other, and any center running both as targets has already found that out. The behaviors that end an issue on the first call cost seconds, like reading a callback number back digit by digit or confirming out loud that the thing is handled, and those go first when average handle time gets squeezed.
A resolution target with no handle time guardrail fails the other way, keeping callers on the line while an agent works a problem that belongs to a different department. Hold one as the target and the other as a band.
What good looks like
Published rates are easy to find and hard to use, since each carries the window and denominator it was measured under. A figure from a center counting at 24 hours on eligible calls says nothing about your floor at seven days on everything inbound, so we went looking for one worth printing here and decided against handing you a number with none of that attached.
So the baseline is yours. Measure three consecutive months without touching the definition, take the median, and hold the floor to improvement against it. A rate tied to agent pay or ranking moves the disposition codes long before it moves the calls, which is the argument for measuring from platform data agents never touch.
What to do when the number is bad
A low rate arrives as a pile of unrelated problems, and sorting it beats coaching it. Pull twenty pairs where a call was followed by a callback, then read the second call first.
Most land in two groups. There is the permissions version, where the agent knew the answer and needed an approval from somebody at lunch, and the knowledge version, where the answer lives in a document nobody can find while a caller waits. A third pile usually sits there too, the calls that were never closing on the phone, and those belong in your eligibility rule.
Coaching only moves the knowledge group, which makes the conversation after the score more useful once somebody has done the sort.
Why a recording can partly verify this one
Most contact center metrics are unauditable after the fact. Handle time is whatever the platform counted, and service level is arithmetic on timestamps. Resolution is different, because a fair share of the evidence sits in the audio.
A transcript shows whether the agent confirmed the issue was handled before the call ended and whether a next step with a date got said out loud. Those lines are rule based, which is the part of automated scoring that holds up, while reading tone or deciding whether a caller sounded satisfied stays shakier, which Aircall and StackAI both note.
Audio cannot tell you whether the fix held or whether the customer emailed two days later. Put it next to the repeat contact number and you get both halves, what the agent did and what the customer did next.
Reading the resolution rate next to your QA scores
A resolution rate climbing while QA scores fall is the pattern worth stopping the meeting for. It usually means agents are closing calls the caller does not consider closed. The disposition code says handled, the scorecard says the confirmation step never happened, and both describe the same call from different ends.
QA score carries its own published orientation, where SQM Group puts the industry average at 85% and treats 90 to 99% as good, and it sits beside coverage rate and critical fail rate among the metrics a small QA program runs. Resolution rate belongs on the operations side of that split, coming off the phone platform rather than out of somebody reading a call, and the two read together catch what neither catches alone.
Checking it takes a batch rather than an argument. Take last month's calls that were coded resolved and produced a callback inside seven days, score them against your existing scorecard, and read how the confirmation and next step criteria came back. One pile holds calls where the card already covers a step the agent skipped, which is a coaching problem with evidence attached. The rest are calls where the card says nothing about why the caller came back, and that missing criterion is the more valuable finding.
That check only works if the scoring was not a sample, so scoring the month instead of three calls per agent is what makes a pile that size readable. A call review scores a batch of your own recordings against your existing scorecard, and pointing it at the calls that produced a callback beats another meeting about whose definition of resolved is right.
Frequently asked questions
What is the first call resolution formula?
Divide calls resolved on the first contact by calls handled in the same window, then multiply by 100. The arithmetic is settled and the two inputs are not, so decide what counts as resolved and which calls enter the denominator first.
What is a good first call resolution rate?
A borrowed rate is not usable as a target, since every published figure carries the window and denominator behind it. Measure three months on a fixed definition, take the median, and improve against that.
What is the difference between first call resolution and first contact resolution?
First call resolution counts phone calls only. The contact version counts whatever channel the customer used, so a caller who phones and then emails about one problem reads as a failure there and a success in the call only version.
How do you measure first call resolution without a survey?
Match inbound contacts against a repeat window, seven days being the common choice, and count a call unresolved when the same customer comes back about the same issue. Matching on the customer record rather than the phone number keeps a borrowed phone out of the numerator.
Does a first call resolution target hurt average handle time?
It raises handle time, and that is usually the right trade, since the steps that end an issue on the first call take seconds and get cut first under time pressure. Set one of the two as the target and hold the other inside a band.
Sources
- SQM Group on the 85% average QA score and the 90 to 99% good band. Third party, unverified.
- Aircall and StackAI on rule based criteria scoring more reliably than tone. Third party, unverified.
- Trinsic feature sheet and call notes, 2026-08-27 and 2026-09-03. Verified internally.
- The 10,000 call worked example, the denominator table, and the seven day window. Editorial judgment.