Short answer: To QA bilingual and Spanish calls, use one QA form for both languages, score each call in the language it was spoken instead of translating it first, give reviewers a pass and fail example in each language, calibrate each language separately, and deliver coaching in the agent's own language.
Why does bilingual call center QA go wrong?
Most QA programs were built for English calls. Spanish calls get pushed to the one bilingual analyst on the team, scored late, or translated and scored as if they were English. Each of those creates a gap: fewer Spanish calls get reviewed, and the ones that do get scored by a different standard.
Bilingual callers also switch languages mid-call. A caller might verify their account in English and explain the problem in Spanish. If your process assumes one language per agent, those calls are where scoring errors pile up.
Should you translate calls before scoring them?
No. A translated transcript has already lost tone, politeness markers and the exact wording a compliance line depends on. "Did the agent read the disclosure?" is hard to judge from a paraphrase. Score the original call and translate only for a manager who needs to read it.
Do you need a separate QA form for Spanish?
Usually not. Two forms drift apart and make English and Spanish scores impossible to compare. Keep one form and make each criterion clear in both languages:
- Write each line as one behavior with a yes or no answer.
- Add a pass example and a fail example in English and in Spanish.
- Say when a line is not applicable, so agents aren't failed for something the call never needed.
- Keep critical fails, such as a missed required disclosure, separate from the weighted score.
- Prefer behaviors ("stated the next step and timeframe") over tone labels ("was friendly"), since tone reads differently across languages and regions.
Not sure your form is ready? Run it through the QA Form Analyzer, and see our common call center QA mistakes.
How do you calibrate QA across languages?
Calibrate each language on its own. Pull 10 to 20 calls per language that a bilingual analyst has already scored, have a second scorer (a person or an AI) score the same calls, and compare line by line. Where scores disagree, the criterion usually needs a clearer definition or a better example in that language.
Spanish isn't one accent, either. If most of your callers are Mexican, Puerto Rican or Central American, calibrate on recordings from those callers rather than a generic sample.
How many Spanish calls should QA review?
At least the same share as English calls. If Spanish calls are only scored when the bilingual analyst has time, those agents get less feedback and their scores swing on a handful of calls. For sample size guidance, see how many QA evaluations per agent.
How should you coach bilingual agents?
Coach in the agent's language, tie each note to the moment in the call, and do it soon after the call. A coaching note in the wrong language, or weeks late, rarely changes behavior.
Can AI score Spanish and bilingual calls?
Yes. AI QA removes the bottleneck of waiting on one bilingual analyst. OttoQA scores calls in any language against your own QA form, with 96% accuracy. It connects to any CCaaS or cloud CRM via API, and setup takes about 2 weeks, including calibration on 10 to 20 calls you've already scored. OttoQA is PCI compliant and signs HIPAA BAAs; SOC 2 is in process. Read how AI-powered auto QA works or the ultimate guide to call center QA.
Frequently asked questions
Should I translate Spanish calls into English before QA scoring?
No. Translation loses tone, hedging and exact wording. Score the call in the language it was spoken, and translate only if a manager needs to read it later.
Do I need a separate QA form for Spanish calls?
Usually not. Keep one QA form with the same criteria, and add a pass and fail example in each language so reviewers read each line the same way.
How do you handle calls that switch between English and Spanish?
Detect the language per call, not per agent, and score the whole conversation as it happened. Test your QA setup on real code-switched calls before you trust it.
How do you calibrate QA for bilingual calls?
Pull 10 to 20 calls per language that a bilingual QA analyst has already scored, compare them with the second scorer or the AI, and fix the criteria where scores disagree.
Can AI score Spanish calls?
Yes. OttoQA scores calls in any language against your own QA form, with 96% accuracy, and setup takes about 2 weeks.
About the author
Tom Laird is the founder of OttoQA and CEO of Expivia, with about 30 years in contact centers. He hosts the Advice from a Call Center Geek podcast. Connect with Tom on LinkedIn.