Last updated: September 9, 2026
What an answering service scorecard measures
An answering service scorecard grades how an operator handled a call against the account script it was answered under. Criteria cover the client greeting, caller capture, urgency triage, message accuracy, and script adherence. The template below runs to 100 points across 16 criteria, sized for a call that ends in under a minute.
Most contact center scorecards assume the agent works for one employer with two or three minutes to get the call right. An operator on an answering service floor picks up in a client's business name, works from that account's script, and is done in about 45 seconds. The next call arrives under a different account with its own greeting, urgency definitions, and on-call schedule, and a busy board carries several hundred of those.
What comes out of those 45 seconds is a message somebody reads at 6 a.m. before their first patient of the day. A callback number off by one digit turns a flawless greeting into a wasted message, which is where this template parts ways with the general call center QA scorecard.
The 100 point answering service scorecard
Sixteen criteria across six sections, weighted by what a mistake costs the account. Section totals are in parentheses.
| Section | Criterion | Points |
|---|---|---|
| Account greeting (15) | Answered with the client's exact business name from the account script | 8 |
| Account greeting | Reached the greeting inside the account's answer time standard | 4 |
| Account greeting | Identified as the answering service only where the account script requires it | 3 |
| Caller identification and capture (20) | Captured the caller's full name and confirmed the spelling | 7 |
| Caller identification and capture | Read the callback number back to the caller digit by digit | 8 |
| Caller identification and capture | Established the caller's relationship to the account, such as patient or tenant | 5 |
| Urgency triage (20) | Classified the call using the account's own urgency definitions | 8 |
| Urgency triage | Asked the account's required triage question before dispatching or taking a message | 7 |
| Urgency triage | Followed the on-call instructions in force at that time of day | 5 |
| Message accuracy (25) | Message reason matches what the caller actually said | 10 |
| Message accuracy | Every field the account requires is present in the message | 8 |
| Message accuracy | No detail appears in the message that the caller did not provide | 7 |
| Script adherence (10) | Delivered the account's required script lines in the order the script sets | 5 |
| Script adherence | Stayed inside the account's do-not-say list | 5 |
| Handling and close (10) | Held the caller within the account's hold standard, with a return check | 5 |
| Handling and close | Confirmed the message back to the caller and closed with the account's approved wording | 5 |
| Total | 16 criteria | 100 |
Grab the CSV below and edit the wording account by account.
Auto fail items
Six failures void the score no matter how the rest of the call went. A critical fail lands on a supervisor's desk the same day, ahead of any monthly average.
| Auto fail | Why it is absolute |
|---|---|
| Answered with the wrong client's business name | The caller believes they reached a different business |
| Recorded a callback number that does not match what the caller said | The message cannot be returned |
| Failed to dispatch a call the account defines as an emergency | The account's escalation instructions were broken |
| Disclosed one account's information while handling another account's call | Confidentiality broken between two accounts |
| Discussed patient information with a caller whose identity was not established under the account's instructions | The service handles that information as a business associate |
| Disconnected a live caller before the message was complete | The interaction produced nothing |
Why message accuracy carries the most weight
A quarter of the points sit in message accuracy, which is our editorial judgment rather than a published standard. The message is what your client actually bought. A greeting delivered exactly as written, attached to a reason the caller never gave, still sends somebody back to the wrong person about the wrong problem.
Answer time looks like it should rank higher, and it is worth 4 points. Weight it heavily and you get a floor that picks up fast and takes the wrong message. It stays because clients ask about it, sitting in their service agreement next to the metrics answering services track.
Scoring a call that lasts 40 seconds
Sixteen criteria is unscorable by hand at answering service volume. Say you run 22 operators and the floor takes 38,000 calls a month. Two scored calls per operator per week is about 190 calls, half a percent of the month, and 13 hours of supervisor time at four minutes each. A failing account rarely announces itself in a sample that thin.
So the criteria here are written to be checked against a transcript. The greeting criterion compares what the operator said against the exact greeting string stored for that account, and do-not-say terms live in a vocabulary library next to the required triage question.
Rule based criteria score reliably, and anything needing a read on tone gets shakier, which Aircall and StackAI both note about automated QA. Every flag carries a confidence score and a timestamp into the transcript, so a supervisor spends the time on the low confidence ones.
Scorecards per account and the client report
One operator gets scored against whichever rubric belongs to the account the call came in under. Same person, same shift, different standard, because the medical office and the plumbing contractor bought different things from you.
That structure moves QA out of the cost column. When a law firm asks how their calls are handled, you answer with the calls taken in their name that month, scored against the script they approved, instead of a promise about your training. CallSniper supports organization branding, so branded client quality reports go out under your logo.
An account that complains every quarter is either right, which the auto fail count shows, or working from instructions nobody can follow.
What belongs in the report you send a client
| Field | What it tells the client |
|---|---|
| Account name | Which of your accounts the report covers, in their own business name |
| Reporting period | The exact window, so it can be compared against the last one |
| Calls scored | How many of their calls the score is built on |
| Average score | Performance against the rubric written for their account |
| Auto fail count | Absolute failures, with the reason attached to each |
| Two example calls | One strong, one weak, with timestamps so they can listen |
The two example calls matter more than the average. A report carrying only the good ones reads as marketing, and the weak call starts the useful conversation about their script.
Handling medical accounts
A service taking calls for a medical practice generally acts as a business associate of a covered entity under a business associate agreement with that practice. The scorecard criterion covers one narrow thing: whether the operator followed that account's identity and disclosure instructions on the call. Your obligations come from your agreements and your counsel rather than from a QA vendor.
CallSniper runs on Trinsic's own model in Trinsic's own cloud, so call audio never reaches a public model. Medical clients ask about that first.
Adapting the template per account
Capture and message accuracy stay fixed across accounts, which is 45 of the 100 points. Six things change per client: the greeting string, the answer time standard, the urgency definitions, the triage question, the on-call rules, and the do-not-say list.
The scorecard grades the operator against the account script and leaves the script itself alone. When an account's triage question is vague, every operator fails that criterion and the fix is a call with the client rather than coaching the floor. Check calibration variance first: two supervisors score the same 10 calls separately, then compare.
Download the template
The rubric downloads as a CSV or a spreadsheet, with no form in front of either. A blank version leaves the fields that change per client empty.
Frequently asked questions
How do answering services measure agent quality?
Most services score a sample against the account script. The usual checks are the greeting, capture accuracy, and whether the message matched what the caller said. Scoring per client account is what lets you report quality back to the business whose calls you answer.
What is a good message accuracy rate for an answering service?
No published benchmark held up when we went looking on 2026-09-08, so set the target from your own baseline. Take the median of a month of scored calls and hold each account to improvement against it.
How fast should an answering service answer a call?
Answer time standards get set per account in the client's service agreement, and no industry rule overrides that, so score against whatever the account says.
Does an answering service handling medical calls need to follow HIPAA?
A service taking calls for a covered entity generally acts as a business associate and handles protected information under a business associate agreement. Confirm your obligations with your own counsel, since the arrangement depends on the contract.
Can this scorecard work with Amtelco or Startel?
Yes. The scorecard does not care what platform produced the audio, and you can hand over the recordings straight out of whatever your system exports. A direct platform connection is a separate conversation, so nothing here waits on an integration project.
A month of account calls, scored
By our own count there are 100 to 150 family owned services of this size in the country, small enough that we have probably already met at ATSI, NAEO, Team SNUG, or CAM-X. That is why QA for telephone answering services gets built differently here than for a 900 seat contact center.
Pick the two accounts that generate the most complaints. The paid pilot scores six months of your own historical audio for those accounts against the scripts you approved, then your supervisors spend about 30 minutes on Zoom with us looking at operator calls under those account names, and the whole thing costs a fixed pilot price. A demo is the shorter version, and a sample client report on one account is smaller still.
Sources
- Trinsic feature sheet and call notes, 2026-08-27 and 2026-09-03. Verified internally.
- 100 to 150 family owned US operators. Client estimate, unverified.
- Amtelco and Startel as platform chokepoints. Client assessment, unverified.
- ATSI, NAEO, Team SNUG, and CAM-X presence. Verified from call notes.
- QATSI, the ATSI quality program (atsi.org). Footprint unverified: Cloudflare challenge on 2026-09-08.
- Aircall and StackAI on rule based scoring holding up better than tone scoring. Third party, unverified.
- The 25 point message accuracy weighting. Editorial judgment for this template.