The Voice QA module
Every call, scored and structured.
Every recording that reaches Fineguide is transcribed and scored against your rules, with the lines from the call that justify each mark. Stop sampling ten calls a week.
How it works
Five steps. Every call.
A QA manager who listens to ten calls a day sees a fraction of what happens on the phone. Voice QA listens to all of them, applies the same rules every time, and has the feedback ready while the call still matters.
- 01 Calls arrive by upload, through the API, or automatically from a connected carrier.
- 02 The AI transcribes the call and separates the two sides, so you read a dialogue between agent and customer rather than a wall of text. It detects the spoken language itself, even when the call switches language halfway through.
- 03 The call is scored against the rules of its department. The AI first works out which categories of rule the call is even about, so a billing question is not marked down for missing a sales rule.
- 04 Each rule comes back with a score, the reasoning behind it, three to five direct quotes from the call, and a suggestion for the agent.
- 05 Managers see the score by department and by agent, which rules pass and which fail, and how all of it moves week to week.
Getting calls in
Upload them, post them, or let them arrive.
Voice QA scores recordings. How they reach the platform is up to you, and nothing is processed, or charged, until you ask for it.
- Upload
- Drop in one recording or a batch, up to 50 MB each. Storing, transcribing and scoring are three separate choices, so a stray click cannot spend credits.
- API
- Post a recording, a link to one, or a transcript you already have. Read back scores, transcripts and dashboard figures the same way.
- Connected carriers
- Moldcell, Orange Moldova and the Fineguide call recording app for Android deliver new calls on their own, checked every minute.
- Automatic scoring
- Turn it on per department and every arriving call is scored without anyone opening it. The department needs at least one rule before anything is scored.
- Numbers you exclude
- Exclude a number, an internal one for example, and its calls are never processed and never appear in the views.
Configuration
Per-department rules, your scoring model.
Every team has its own quality bar. Voice QA lets you write it down, with different rules and weights per department, and one engine handling sales calls, support calls and retention calls side by side.
Departments
Organise by team, location or function. Each department has its own rules, its own settings and its own dashboards.
Categories and rules
Write what good sounds like, such as the opening, discovery, objection handling and compliance, grouped into categories. Up to 50 rules per department.
Fixed or weighted
Score every rule out of the same number, or give each rule its own weight. The final percentage follows your weights, not ours.
Rules that do not apply
A rule the call never had the chance to trigger is marked not applicable and left out of the total, instead of scored zero. Rules can also apply to incoming calls only, outgoing only, or both.
Agent scorecards
The scorecard is ready as soon as the call is scored, with the quotes behind every mark and a suggestion per rule, and it can be emailed to the agent automatically.
The customer record
Calls match to a contact by phone number and land on the same customer timeline as chats and tickets. The customer is one story, not three.
On every call
The score, and the reason for it.
A number on its own starts an argument. Every mark Voice QA gives is backed by the words that earned it, so a coaching conversation begins with evidence instead of opinion.
- The transcript
- Both sides of the call, separated, with the automated IVR introduction stripped out.
- Rule by rule
- A score, the reasoning, three to five direct quotes, and one concrete suggestion. Quotes stay in the language the customer used, even when the report is written in another.
- How the customer felt
- Their mood tracked across the call rather than summed up at the end, a satisfaction reading, and what they called about.
- Fields you define
- Tell the AI what else to capture on every call, such as the product mentioned, the complaint type or a competitor named, as text, a number, yes or no, or a pick list. Caller name, company, phone and email come as standard.
- The recording
- Playable next to the score, so you can hear the moment yourself.
- The report language
- Conclusions, reasoning and suggestions come back in the language you choose, from 30, whatever language the call was in.
Cases and dashboards
One call is an incident. Ten is a pattern.
Individual scores tell an agent how they did. The views above them tell you where the operation is losing customers.
- Cases
- Group up to ten calls about the same problem for the same customer and score the case as a whole: whether it was resolved, how long it took, how many people the customer had to deal with, and how their mood moved from the first call to the last.
- By department and agent
- Average score, how it compares with last month, and who is furthest ahead and furthest behind.
- By rule
- Which rules your team passes and which they keep failing, so training goes where it is needed.
- Over time
- Score, minutes processed and call volume, day by day.
- Out to your systems
- Every finished call can be pushed to your own webhook, or to AmoCRM or Kommo.
Scoring is charged per minute of audio, with a one-minute minimum per call. Transcribing without scoring costs half.
Stop sampling. Start scoring.
See Voice QA running on your own recordings in a 15-minute walkthrough.