Build a weighted QA (quality assurance) form for one contact type, check that it produces scores you can defend, and calibrate your evaluators blind: each scores the same calls alone, and nothing is compared until everyone has scored.
Every rule, cut point and source is published in the QA method.
1. The form
Each criterion needs a definition of what earns a yes and a tag for what it measures. An auto-fail criterion, one whose miss fails the whole evaluation, also needs the reason it must be one.
2. Form check
What this form measures, by share of its weight. It is shown as a fact only, because no source sets a right mix.
42%
Customer outcome
12%
Compliance
47%
Process and behavior
The form passes every published check.
3. Score a call
For each evaluator in a calibration session: enter your initials and the call ID, mark every criterion, then send your submission code to your QA lead. You see only your own score. Nobody sees how it compares until every evaluator has scored every call.
Opening + Authentication (10%)
Proper greeting and identificationAgent greets the customer and gives their own name and the company name in the opening.
Auto-fail:Customer verified per policyAgent completes every identity check the verification policy requires before discussing or changing the account.
Active Listening + Discovery (20%)
Acknowledged customer concernAgent names or restates the customer's concern before moving to solve it.
Asked clarifying questionsAgent asks at least one question that narrows the issue before proposing a fix, unless the customer has already stated it in full.
Restated issue to confirm understandingAgent summarizes the issue back and the customer confirms it.
Knowledge + Accuracy (25%)
Auto-fail:Provided correct informationEvery fact, policy or instruction the agent gives is correct against the current knowledge base.
Used appropriate resourcesAgent uses the knowledge base articles or tools the process names for this contact type.
Auto-fail:Did not guess or provide unverified infoAgent states nothing as fact that they have not confirmed; when unsure, they say so and check.
Resolution + Ownership (25%)
Resolved the issue or set clear next stepsThe issue is resolved on the contact, or the customer leaves knowing the next step, who owns it and when it happens.
Took ownership (no unnecessary transfers)Agent resolves what is within their authority and transfers only when the process requires it.
Set expectations for follow-upWhere follow-up is needed, agent states what happens next and by when.
Closing + Compliance (20%)
Summarized resolution and next stepsAgent recaps what was done and what happens next before closing.
Auto-fail:Required disclosures were providedEvery disclosure the script or regulation requires for this contact is given in full.
Professional closingAgent checks whether anything else is needed and closes courteously.
0 of 14 criteria marked. Your score and submission code appear once every criterion is marked.
4. Calibration session
For the QA lead. A session runs in four steps:
Pick 3 or more calls that range from weak to strong. The method measures whether evaluators tell good calls from weak ones, so a set of similar calls reads as low agreement.
Send each evaluator this form: . It opens this form at step 3, with nothing marked.
Each evaluator enters their initials and the call ID, marks every criterion, and sends you the code that appears. One code per evaluator per call.
Paste every code below, one per line. Results stay sealed until every evaluator has scored every call.
To see how results read first, : three evaluators on three calls, on the General Inbound form.
0 codes accepted from 0 evaluators on 0 calls.
5. Findings and next step
0 critical0 high0 medium0 note
Next step: Run a blind calibration session: have at least two evaluators score the same calls with this form, then paste their codes below.
One form per contact type. A password reset and a billing dispute call for different criteria. A single form for every contact type ends up too generic for complex calls, or it marks simple calls down for criteria that do not apply. Build a form for each contact type, complexity or risk level, and weight what matters for each.
Take this with you
The report downloads immediately. No email required, no wall.
This scenario is too detailed to fit in a link. Download instead.
A person on our team sends it, usually within one business day. Your address is used to send this report and to reply if you ask a question. Nothing you entered leaves your browser unless you send it here, and the download never asks for it.
Run this next
Each result raises a sharper question. This is the diagnostic that answers it.
Send this analysis for an independent review. You get a written read on what the numbers support, what they do not, and which diagnostic is worth running next. No sales call. No vendor introduction.
Next step
Calibrate
Calibrate
14 criteria, 0 form findings. Calibration: not run.