On this page
- Evidence snapshot
- Define the language work before testing anyone
- Test the tasks the queue actually requires
- Data and decision boundary
- Score comprehension separately from accent
- Use two raters and one rubric
- Design the test around the law and the job
- Give candidates a fair practice window
- Copy-ready call and handoff lines
- Recheck language readiness after a script change
- Questions managers ask
- Keep planning
- Sources
A language test is useful when it mirrors the queue's real tasks, scores comprehension and accuracy separately, and applies the same standard to every candidate.
Evidence snapshot
Define the language work before testing anyone
Start with the tasks the queue actually requires in the second language. A billing queue may need numbers, dates, and payment terms; a retention queue may need empathy and objection handling. Write the task list before choosing a test. A general conversation score does not prove that an agent can explain a policy or read a disclosure accurately. The test should mirror the work, not a language classroom.
Decide which tasks are required on day one and which can be learned with support. Mark the required tasks with a clear standard, such as complete the task without repeating the customer or escalating for language reasons. Share the standard with the recruiter and the trainer. If the standard is vague, every interviewer will apply a different bar and the result will not be comparable.
Test the tasks the queue actually requires
Use short role-plays built from real, de-identified call reasons. Ask the candidate to explain a policy, confirm a date, handle a customer who disagrees, and close with a next step. Score the outcome and the clarity, not the accent. A candidate who solves the customer's problem in plain language is stronger than one who sounds native but misses the request. Keep the scenarios at the same difficulty across candidates.
Include one listening task with background noise if the queue is noisy. Include one reading task if the agent must read a disclosure or a knowledge article. Include one writing task if the agent writes customer notes. The test should cover every channel the role uses. A phone-only test cannot prove that the agent can write a note the next person can understand.
Data and decision boundary
Use this table as a starting point, then match each row to the client's tools and call guide. The manager column stays outside the team member's normal authority.
| Data or request | Team member can | Manager keeps |
|---|---|---|
| Language screen | Complete the approved role-play | Approve the standard and retest |
| Scoring | Score each dimension separately | Resolve rater disagreement |
| Accommodation | Offer the approved format | Approve the accommodation |
| Reassignment | Record the language gap | Decide training or reassignment |
Score comprehension separately from accent
Separate the rubric into comprehension, clarity, accuracy, and task completion. Accent is not a scored dimension because it does not predict whether a customer understands the agent. Score whether the candidate understood the question, answered the actual request, and used words the customer could follow. Record examples for each score so a second rater can review the decision. A single overall impression hides the reason for a pass or fail.
Give the candidate a chance to repeat or rephrase, just as an agent would on a live call. Score the recovery, not only the first answer. A candidate who notices confusion and repairs it is demonstrating a useful call-center skill. Do not penalize a regional pronunciation that customers routinely understand. The question is whether the work gets done clearly, not whether the accent matches a recording.
Use two raters and one rubric
Have two trained raters score the same recorded role-play independently. Compare the scores and discuss the gaps before making a hiring decision. If the raters disagree on more than one dimension, the rubric is probably ambiguous. Revise the example or the definition and score another sample. Two raters are not a guarantee of fairness, but they make a single interviewer's preference visible.
Calibrate the raters on the same set of candidates before the live round. Include one strong, one borderline, and one weak example. Write down why each example received its score. Refresh the calibration on a fixed schedule and whenever the queue, script, or customer base changes, then review the results together. A rubric that is never calibrated drifts toward the raters' personal preferences.
Design the test around the law and the job
Base every language requirement on the duties of the role, not on customer preference or national origin. The EEOC explains that fluency requirements and English-only rules must be job-related and consistent with business necessity. A blanket English-only rule is not the same as a requirement to perform the job in a particular language. Have the client and employment counsel review the standard before it is used.
Apply the same test to every candidate for the role and document the result. Offer the test in a consistent format and provide reasonable accommodation when needed. Do not use a language test to screen out a protected group or to enforce a preference unrelated to the work. Keep the scoring record with the hiring file and review the pass rate by group over time.
Give candidates a fair practice window
Tell candidates the format, the tasks, and the standard before the test. Offer a short practice scenario so they can hear the expected flow. A candidate who understands the process is more likely to show their actual ability. Do not surprise them with a hostile customer or an unfamiliar system unless the role requires that exact condition and the practice covers it.
Allow a reasonable retest window for a candidate who narrowly misses one dimension. Use the same rubric and a different scenario to avoid memorization. Record the retest result and the reason. A retest is not a lowering of the standard; it is a check on whether the first result was stable. If the candidate fails the same dimension twice, the evidence is stronger.
Copy-ready call and handoff lines
Explaining the test
You will handle three short calls in Spanish: a billing question, a date change, and a customer who disagrees. We score comprehension, clarity, accuracy, and task completion, not accent.
Handling a borderline result
Your task completion was strong, but the date confirmation was unclear. We will retest that one task with a different scenario before making a decision.
Recording the decision
The candidate passed comprehension, clarity, and accuracy, and missed one listening repair. The second rater agreed, and the retest is scheduled for Thursday.
Recheck language readiness after a script change
A script change can introduce new terms, disclosures, or customer objections that the original test never covered. Review the change and decide whether it requires a new language check. Retest only the affected dimension and keep the result with the training record. An agent who passed last year may not be ready for a new payment disclosure or a new product name.
Use live call samples to confirm that the tested skills appear in the work. Listen for comprehension repairs, accurate numbers, and clear next steps. If the samples show a gap, update the training before adding more volume. Do not assume that a passing score at hire predicts performance after a major change. Review at least five samples per agent before drawing a conclusion.
Questions managers ask
Is a language test legal?
A language requirement can be lawful when it is job-related and consistent with business necessity, applied consistently, and reviewed by employment counsel.
Should accent be scored?
No. Score whether the customer can understand the agent and whether the task is completed accurately.
When should an agent be retested?
After a major script, product, or disclosure change that introduces language the original test did not cover.
Sources
- Occupational Outlook Handbook: Customer Service RepresentativesBureau of Labor Statistics, 2026. Employment, wage, training, and outlook figures for customer service representatives.Source 1
- National Origin DiscriminationU.S. Equal Employment Opportunity Commission, 2026. EEOC guidance on national origin, language, and employment decisions.Source 2
- English-Only Rules and National Origin DiscriminationU.S. Equal Employment Opportunity Commission, 2026. EEOC guidance explaining that language requirements must be job-related and consistent with business necessity.Source 3
- Fact Sheet #64: Call Centers under the Fair Labor Standards ActU.S. Department of Labor, Wage and Hour Division, 2008. Hours-worked, rest-period, and recordkeeping rules for call center employees.Source 4
