How do you run call QA for a bilingual sales team?
Run one scorecard, not two, and score every call in the language it was spoken in. A bilingual floor breaks most call QA tools not at transcription but at analysis: the vendor transcribes Spanish and then runs its scoring, alerting and coaching features on English only, so half your calls come back as text nobody grades.
Why "we support Spanish" doesn't answer the question
Because it is a claim about transcription, and QA happens after transcription.
The clearest published example is Gong, and it comes from Gong. Their language page lists "Spanish (USA, Spain, all countries)" as supported — that part is true. Four paragraphs down, the same document says: "Some Gong features only generate results for calls in English. For calls in other languages, these are not available:" and then lists email labels, red flag alerts, Gong-recommended to-dos, and example-based smart trackers (Gong help centre, checked 18 September 2026).
Read that as a buyer with a bilingual floor. The transcript arrives. The alert that tells a supervisor something went wrong on the call does not. You pay the same seat price and get a searchable transcript instead of a reviewed call.
This is not a criticism of Gong specifically — they are unusually honest in documenting it, which is why they are the example we can cite. The pattern is the category's. Before you buy anything, the question is not "do you support Spanish". It is "which features stop working on a Spanish call, and can I see that in writing?"
The two-scorecard trap
The workaround most agencies land on is running a separate process for Spanish: a bilingual supervisor reviews those calls by hand, or a second rubric gets written for them.
Both fail for the same reason. Two rubrics produce two scales, and two scales cannot be compared. When the Spanish-speaking agent scores 78 and the English-speaking agent scores 84, nobody in the room can say whether that is a performance gap or a rubric gap — so the number stops driving decisions and the whole exercise turns into paperwork. The manual review fails faster: a supervisor sampling by hand gets through about 4 calls per agent per month, which on a floor making 200 dials per agent is a 2% sample. A 2% sample tells you almost nothing about the 98%, in either language.
One rubric, applied to every call, is what makes a per-agent number mean something.
How it actually works when it works
- The campaign script is the rubric. Not a generic QA form — the script your agents were trained on, in both languages. The same requirements, expressed in each language, not translated at scoring time.
- The language is detected per call, not per agent. Real bilingual agents switch mid-call, or open in English and finish in Spanish after the prospect relaxes. A per-agent language setting gets that wrong the moment it matters.
- Scoring happens in the language spoken. Translating first and grading the English is where meaning goes: the difference between a firm commitment and a polite deferral in Spanish rarely survives a machine translation, and that difference is the appointment.
- The coaching note goes to the agent in their language, the same day. A note in the wrong language gets read once.
What to ask a vendor before you buy
| Ask this | Why it separates the field |
|---|---|
| Which features do not run on a non-English call? Show me the doc. | Most vendors answer the language question with a list of transcription languages. Very few publish the feature gap. Gong does; make the others say it in writing. |
| Is language detected per call or set per agent? | Per-agent settings mis-handle exactly the calls a bilingual floor cares about. |
| Does a Spanish call cost more than an English one? | Language add-ons are common outside call QA and are creeping in. Get the answer before the contract, not at the first invoice. |
| Can one rubric cover both languages, or do I maintain two? | Two rubrics means two scales, and two scales cannot be compared. |
| What does the agent receive, and in what language? | QA that the agent cannot read is reporting, not coaching. |
Where LeadProof stands on each of those
LeadProof detects the language per call, scores the call in the language it was spoken in against your campaign script, and charges the same either way — $21–31 per agent per month, published, with 80 audits included per agent. Nothing joins the call: it reads what Zoom Phone and Zoom Meetings already recorded. The coaching note reaches each agent in their language the same day.
Where someone else is the better answer: if what you need is live guidance during the Spanish call rather than a review after it, Balto has been shipping real-time guidance in Spanish since 2022 and we do not do that at all. We compare ourselves to them line by line here.
Frequently asked questions
Does the transcript need to be translated into English first? No, and translating first is where the errors come from. Scoring reads the Spanish, because the specific words that separate a firm appointment from a polite brush-off are the ones a translation flattens. The transcript stays in the language of the call.
What about an agent who switches between English and Spanish in the same call? That is the normal case on a bilingual floor, not the edge case. Detection has to run on the call, not on the agent's profile — an agent flagged "Spanish" whose best calls open in English will be scored against the wrong expectation every time.
Do I need a bilingual supervisor to run this? No. That is the point of a single rubric: the supervisor reads a score and a coaching note built from the same requirements regardless of language, and can act on a Spanish call's result without speaking Spanish. Reading the call itself still requires the language — reviewing the floor does not.
Book the 20 minutes · Pricing · What happens to your recordings