Automatic Call Quality Assurance: Every Call Reviewed, Not Just a Sample
TalkWiz reviews every sales and service call against your organization's criteria and weights, and gives each call a score from 0 to 100 with reasoning and a quote from the call for every finding. Managers stop searching for calls at random and start from the calls that really need a review.
Updated
Why a manual sample is no longer enough
A QA manager listening to a call spends at least the length of the call, plus a few minutes filling in the form. So most call centers review only a small fraction of their calls, sometimes no more than 2 to 3 percent. The result is familiar:
- The monthly score comes down to luck: one good or bad call defines the agent for a whole month.
- Different reviewers, different scores: two managers give the same call different scores, and agents feel it.
- Problems surface late: an angry customer, a compliance statement that was not said or a wrongly marked status stay in calls nobody heard.
What TalkWiz checks on every call
TalkWiz analyses every call it receives, not a sample. Each call is reviewed against the same scorecard, the same way, in under a minute.
- Your scorecard: the criteria, the description of each one and its weight are set up for your organization. Each criterion gets a score from 0 to 10, and the overall call score (0 to 100) is calculated from the weights with a fixed formula, not by the language model.
- Reasoning and a quote for every finding: next to each score there is a short explanation and a quote from the call, so you can check it in seconds without listening to the whole call.
- Fair to the agent: a criterion that was not needed in the call (for example handling objections when no objections came up) gets no score and does not pull the average down.
- Compliance and script: statements the agent must make, such as the recording notice, terms or the right to cancel, are checked on every call: said, partly said or not said. Different wording with the same meaning is recognized too.
- Strengths and tips: up to three things the agent did well, and two or three practical tips based on moments in the call itself.
Want to see what it looks like? The sample analysis shows one call with everything the system produces from it.
When a manager disagrees: manual score, feedback and calibration
Automatic QA is only worth something if people trust it. So the manager always has the last word:
- A manager score next to the AI score: a manager can give any call a manual score. The manual score is kept next to the automatic one without overwriting it, and on the dashboard the manager's score is the one that counts.
- Feedback on every finding: each finding can be marked correct or incorrect. Agents can also mark a finding they think is wrong, which is their way to dispute it.
- Calibration: the organization's admins see on the dashboard how many findings were marked correct in the last 30 days. When there is a steady gap on one criterion, its definition is usually not sharp enough, and you refine it in the scorecard.
How to build a good scorecard and calibrate between reviewers is covered in the guide Call Quality Assurance.
What the manager sees day to day
- A review queue: instead of picking calls at random, the manager gets the calls with a score under 60, a high churn risk or a status the agent marked that does not match what happened in the call, which have not yet had a manager score.
- A team heatmap: agents against criteria. You see at once whether a weakness belongs to one agent or to the whole team, and whether group training is needed.
- Alerts on unusual calls: a low score, a missed compliance statement, an angry customer or a legal threat. The urgent ones are sent straight away, the rest are gathered in a daily summary.
- A smart daily report: at the end of the working day, managers get an email with the points that need attention and the calls worth hearing.
For the metrics worth measuring next to the score, read Call Center KPIs.
How to start
- Connect your calls: CallMarker sends every transcribed call automatically when it ends. You can also upload recordings or files (Excel, CSV, Word), or send calls through the API.
- Set up the scorecard: in settings, under "Scorecard", copy the criteria from your existing QA form, with a description and weight for each, and add the statements the agent must say. No form yet? Start from the default seven-criteria scorecard and refine it later, and we are happy to help set it up.
- Test it on your own calls: the first 100 calls are analyzed free, with no credit card and no time limit.
After the free calls, plans start from ILS 590 a month + VAT, with unlimited users. Details on the pricing page.
Frequently asked questions
Can we keep our own QA form?
Yes. In settings, under "Scorecard", you set the criteria yourselves: what each one checks, the weights and the statements the agent must say. No form yet? Start from a default seven-criteria scorecard and refine it later.
What happens when a manager disagrees with the score?
The manager gives the call a manual score. The manual score is kept next to the AI score without overwriting it, and on the dashboard the manager's score is the one that counts. Each finding can also be marked correct or incorrect, and agents can mark a finding they think is wrong.
How accurate is the analysis in Hebrew?
TalkWiz was built for Hebrew from the start, including slang and transcripts with errors. Rather than promise a number, every finding comes with a quote from the call so you can check it in seconds, and the dashboard shows how many findings were marked correct in the last 30 days. The best test is your own calls: the first 100 are free.
Does a criterion that is not relevant to the call lower the score?
No. A criterion that was not needed in the call, for example handling objections when none came up, gets no score and is left out of the calculation. So a short reminder call is not penalized for having no needs discovery.
Does automatic QA replace the QA manager?
No. It replaces searching and random listening. The manager gets a queue of the calls that need attention, and spends the time on feedback and coaching.