Last updated: September 12, 2026
What a call monitoring form is
A call monitoring form is an observation sheet a supervisor fills in while listening to a call. Each line records whether a behavior happened, with room for a note on what was actually said. Point values stay off it on purpose, so the reviewer describes the call before anybody grades it.
A monitoring form gets marked while the call is running. A scorecard gets filled in afterward against a recording, where the arithmetic is the whole point. Different jobs, so the fields come out different.
A scorecard line asks how many of the 8 points for empathy an agent earned, a judgment nobody makes at the speed of a live call. Whether the agent read the callback number back digit by digit is markable with a pen while the caller is still talking. Hand a supervisor a relabeled scorecard for live use and they hunt for the right box instead of listening.
The monitoring form
Twenty one observation points across five groups. Mark one result per line and write what you heard in the note column. Anything the call type did not call for gets NA rather than a no, since a no you cannot defend later is worse than a blank.
| Section | Observation point | Result | Note |
|---|---|---|---|
| Opening | Answered inside the program's target queue wait | Yes / No / NA | |
| Opening | Delivered the greeting the account script sets, with the agent's own name | Yes / No / NA | |
| Opening | Confirmed who they were speaking to before going further | Yes / No / NA | |
| Opening | Asked what the caller needed before offering a solution | Yes / No / NA | |
| Listening and control | Let the caller finish the first description of the problem without cutting in | Yes / No / NA | |
| Listening and control | Asked a follow up question using the caller's own words | Yes / No / NA | |
| Listening and control | Paraphrased the request back before acting on it | Yes / No / NA | |
| Listening and control | Stayed inside the hold standard and checked back with the caller | Yes / No / NA | |
| Listening and control | Kept their own volume and pace steady when the caller escalated | Yes / No / NA | |
| Information handling | Verified the account through the identity steps the program requires | Yes / No / NA | |
| Information handling | Read the callback number back digit by digit | Yes / No / NA | |
| Information handling | Gave an answer that matches current policy rather than memory | Yes / No / NA | |
| Information handling | Told the caller what happens next and when | Yes / No / NA | |
| Information handling | Documented the call during it rather than in wrap time | Yes / No / NA | |
| Compliance | Delivered the recording or monitoring disclosure before the first substantive exchange | Yes / No / NA | |
| Compliance | Stayed inside approved wording for anything regulated on this account | Yes / No / NA | |
| Compliance | Asked before placing the caller on hold or transferring | Yes / No / NA | |
| Compliance | Avoided a commitment the program does not authorize them to make | Yes / No / NA | |
| Close | Checked whether the caller had anything else outstanding | Yes / No / NA | |
| Close | Restated the agreed next step and who owns it | Yes / No / NA | |
| Close | Closed with the account's approved wording | Yes / No / NA |
Edit the compliance group first. Four lines covers a general inbound queue and covers almost nothing on a Medicare sales floor, where TPMO disclaimer placement and the Scope of Appointment record are their own checks.
The header fields
Seven fields across the top. The last three are the ones people forget.
| Field | What goes in it |
|---|---|
| Agent | Name and team, so the form files itself |
| Date and time | The clock time of the call, not the day you got around to reviewing |
| Queue or account | Which script the agent was working from |
| Call length | Sets the expectation for how much the agent could reasonably cover |
| Monitoring mode | Live, side by side, or recorded |
| Reviewer | The person who marked the form, since a second reviewer may score the same call |
| Recording reference | The call ID or file name, so somebody can hear it later |
Without the recording reference, everything below it turns into one supervisor's memory of a call that nobody can play back.
Download the form
The form downloads as a CSV with no email wall in front of it. Delete the groups your program does not run and rewrite the compliance lines for the accounts you take calls for.
Live monitoring, side by side, and recorded review
Three ways to watch a call, and the choice changes what you learn more than the form does.
Silent monitoring tells you what the floor sounds like when nobody is standing there. Side by side gets the agent talking about why they made a choice, which no recording will tell you, at the cost of an agent who behaves slightly better with a supervisor at their shoulder. Recorded review is the slowest and the only one where a second person can check your work.
Most programs start silent, then drift toward recorded review once the first disagreement about what an agent said turns into two people remembering different calls.
What the law expects before you monitor
Two separate questions live here and they get mixed up constantly.
Consent to the recording or monitoring turns on the jurisdictions the call touches. Some require one party to agree and others require everybody on the line, so a call crossing a state border can pull in more than one rule. Notice to your own employee is a separate matter, running on your employment agreements and internal policy.
A QA vendor should not be answering either one for you. Consent and disclosure rules sit on the compliance checklist, with the disclosure line itself written out in the versions that go into a greeting.
Writing the note so it survives being read back
The note column is where a monitoring form earns its place and where most of them go wrong. Describe the event rather than the trait. "Agent was dismissive" starts an argument about a person's character, while "at 2:14 the caller asked twice about the cancellation fee and got the same scripted line both times" starts a conversation about thirty seconds of a call.
Two notes on what went well and one on what did not is a fair ceiling for one call. Past that, the agent reads the form as a list of everything wrong with them.
Put a clock time on every note, live monitoring included. A number delivered on its own gets disputed, and the same number delivered with the thirty seconds behind it gets discussed, which is the whole case for playing back the deduction alongside the form.
Turning observations into a score
Three or four rounds of monitoring is usually enough to see the pattern. Pull the notes together, find the observations that keep showing up, and rewrite each one as a criterion naming an observable event rather than a quality. Then decide what a failure on it costs, and weight it from there.
Weighting is a real design problem with its own page. How to build a call scoring rubric covers scale choice and what to do with a criterion two reviewers disagree on every single time, while the weighted QA scorecard runs to 100 points across 17 criteria if you would rather start from something already argued over.
Monitoring first and scoring second is the order that works, though plenty of programs run it backwards and survive on a rubric built from what somebody imagined their calls contain.
What a form cannot see
A monitoring form records what one reviewer noticed, and two supervisors on the same call mark different lines while both feel confident about it. Calibration variance measures that gap, and calibrating two reviewers is how you find yours before a dispute does.
Automated flagging carries the same shape of limit somewhere else. Aircall and StackAI both describe machine scoring as reliable on rule based criteria and shakier on tone, which matches what we see: a required phrase either appeared in the transcript or it did not, while whether an agent sounded patient is a call a model should not be making alone. CallSniper attaches a timestamp and a confidence score to every scored line so supervisors spend their time on the low confidence ones.
Speaker separation earns a plain warning too. Any criterion depending on who said a line inherits whatever the diarization got right, and audio arriving as one mixed channel is harder to split than audio recorded with each party on its own leg.
The acknowledgment line and the dispute window
A form the agent never signs changes nothing, so the bottom block matters as much as the lines above it. Four fields close the document out.
| Field | What it does |
|---|---|
| Agent acknowledgment | Signature or a click confirming the agent received and read the form |
| Acknowledgment date | The date the agent saw it, which starts the clock |
| Dispute by | The last date the agent can challenge an observation |
| Where a dispute goes | A named person other than the reviewer who wrote the form |
Acknowledgment means received, never agreed. Say so on the form in one line, because an agent who reads a signature as consent will either refuse to sign or sign resentfully.
Five working days from the acknowledgment date is our recommendation, and the number is editorial judgment rather than a standard. The test is whether the window survives a normal schedule. An agent on four ten hour shifts who acknowledges a form on Friday needs it still open when they next work, with room to listen to the call rather than argue from memory.
Route disputes to somebody other than the person who wrote the form, since asking a reviewer to overturn their own observation puts them somewhere nobody handles well. A program with one QA lead can send it to the operations manager or a peer supervisor. A resolved dispute then goes back to the rubric change log, because a criterion that produces the same fight twice is badly written rather than a discipline problem.
A paid pilot runs six months of your own recordings through these observation points and the scorecard that follows them, uploaded by hand with no telephony project attached, for a fixed pilot price.
One mechanical detail trips programs up: date the form the day you deliver it rather than the day of the call, since the window runs from delivery and a form sitting in a supervisor's drafts has already eaten most of it.
Frequently asked questions
What should be included in a call monitoring form?
Seven header fields at the top, including the monitoring mode and a recording reference, then observation points grouped by opening, listening and control, information handling, compliance, and close. Every line takes a yes, a no, or a not applicable, plus a note. Point values stay off the form.
What is the difference between a call monitoring form and a QA scorecard?
They do different jobs at different moments. A monitoring form is marked while the call runs, and the weighted QA scorecard is scored against a recording afterward, with weights, a total, and an auto fail list.
Can a call monitoring form be used for live calls?
Yes, that is what it was built for. All three monitoring modes work off the same form, though only recorded review lets a second reviewer check the first one's marks.
How often should supervisors monitor calls?
SQM Group reports that 60% of call centers evaluate five or more calls per agent per month. Treat that as a floor rather than a target, since the right number moves with call volume and how much regulated content the queue carries.
Do you have to tell agents they are being monitored?
Employee notice and caller consent are separate questions. Notice to your agents runs through your employment agreements and internal policy, while consent to the recording depends on the jurisdictions the call touches, which the compliance checklist carries. Run both past your own counsel rather than treating any of this as legal advice.
Does an agent have to sign a call monitoring form?
That depends on your program and your employment policy rather than any rule we can point you at. Ask for acknowledgment of receipt rather than agreement, and say so on the form itself.
How long should an agent have to dispute an observation?
Five working days from the date they acknowledge the form is what we recommend, labeled as judgment rather than a standard. The window has to be long enough for the agent to work a shift and listen to the recording.
Sources
- 60% of call centers evaluate five or more calls per agent per month. SQM Group. Third party, unverified.
- Machine scoring holding up better on rule based criteria than on tone. Aircall and StackAI. Third party, unverified.
- Consent requirements differing by jurisdiction. State statutes. Unverified here on purpose: cite a dated legal survey rather than a state list.
- Transcript timestamps, per line confidence scores, and manual upload with no telephony integration required to start. Trinsic feature sheet, 2026-08-27. Verified internally.
- The five working day dispute window and the two positive notes per call ceiling. Editorial judgment for this template.