Evala
Back to the Lab

What a Real Communication Assessment Should Measure (And Most Don't)

· 7 min read· CEFR, Contact Center, Assessment
What a Real Communication Assessment Should Measure (And Most Don't)

Language testing answers a 20-year-old hiring question. See the five things a modern contact center assessment should actually measure to predict job success.

For twenty years, the contact center hiring question has been mostly settled. Can the candidate speak the language? Run a test, get a score, move on. That made sense when the job was simpler and the customer's patience was longer.

It does not make sense anymore. The bar has moved, the work has changed, and a lot of the tools used to screen for it are still answering a question from two decades ago.

We have heard this directly from the people who run these operations. Language alone is no longer the thing that separates a good hire from a costly one. It is necessary, but it stopped being sufficient a long time ago. The agents who fail are rarely the ones who could not speak. They are the ones who could speak but could not follow, could not adapt, could not be put in the right seat, or did not have the aptitude the role actually demanded. A test that only checks the language misses all of it.

So the real question is not whether your assessment is strict enough. It is whether it is measuring the right things at all.

Why a language score stopped being enough

Start with what a traditional language test actually tells you. It tells you a candidate knows grammar, has vocabulary, can pronounce words, and can produce fluent speech in a controlled setting. Real, useful, and roughly where the industry has been parked for years.

Here is what it does not tell you. Whether the candidate can follow a customer who changes their request halfway through. Whether they can hold the thread when a second constraint lands. Whether they have the reasoning speed to diagnose a problem while documenting it. Whether they are even in the right kind of role for their strengths. Whether they understand the domain well enough to know when a customer is describing something that does not add up.

None of that is language. All of it is communication and fit, and it is exactly where modern hires succeed or fail.

The data has been pointing here for a while. When Robert Half surveyed 1,400 executives on what causes a hire to fail, the single most cited factor, aside from raw performance, was a poor skills match. Not a language gap. A mismatch between what the person could do and what the seat required. And the cost of getting that wrong is not small. The U.S. Department of Labor estimates a bad hire can run up to 30 percent of that person's first-year earnings, and SHRM puts the typical figure around 40 percent of salary once you fold in the hidden costs. In a high-volume contact center, that is not one expensive mistake. It is a pattern repeating across hundreds of seats.

The five things that actually predict whether someone can do the job

If language is one input rather than the whole answer, what does a complete picture look like? From everything we see on the floor, it comes down to five layers. Most tests measure the first one and stop.

Language proficiency. The baseline. Grammar, vocabulary, pronunciation, fluency. Still necessary, just no longer the finish line. This is the part the industry already knows how to test.

Comprehension. Not whether the candidate can speak, but whether they can understand. Can they catch the actual intent behind what a customer says, including the part the customer did not say clearly? A frustrated caller rarely states their problem in clean, test-shaped sentences. Comprehension is the difference between an agent who solves the stated problem and one who solves the real one.

Real-time conversational handling. This is the right person question. A live interaction is not a single response. It is a moving target where the customer interrupts, redirects, adds pressure, and changes their mind. Some candidates handle the first turn beautifully and fall apart on the follow-up. A multiple-choice test will never see that, because it only ever asks for one answer to one frozen prompt.

Aptitude and role fit. This is the right seat question. Two candidates can have identical language scores and belong in completely different roles. One has the reasoning speed and composure for escalations. The other is better suited to structured, lower-variability work. Logical reasoning, processing speed, and judgment are known predictors of how fast an agent ramps and how well they hold up in high-pressure situations. Putting a strong candidate in the wrong seat produces the same result as hiring a weak one: early exit.

Domain readiness. Whether the candidate can pick up and apply the specific knowledge the program requires. A healthcare queue, a fintech dispute line, and a food-delivery support desk demand different things. Communication in a vacuum is not the same as communication inside a domain with real stakes and real complexity.

A candidate who passes only the first layer can still fail on the other four. That is precisely why so many "passing" hires wash out, and why a single language score gives hiring managers false confidence.

Right Person. Right Seat.

This is the idea behind how we think about Evala, and it is more than a tagline. The right person handles the real conversation in front of them. The right seat matches their aptitude to the work the role actually demands. A hire only sticks when both are true, and a language test on its own can confirm neither.

Where Evala started was the part of this the industry had been getting wrong for years: scoring voice and writing as separate signals, and assessing real, multi-turn conversation instead of recall. That alone exposes things a blended language score hides, like a strong writer whose verbal clarity collapses under pressure.

Where we are heading is the rest of the picture. We are actively building toward an assessment that reads comprehension and intent, gauges aptitude and role fit so candidates land in the right seat rather than just clearing a bar, and accounts for domain readiness. The goal is not a longer test. It is a truer signal, one that answers the question hiring managers are actually asking, which was never "does this person know English" but "can this person do this job, in this seat, on a real call, starting now."

What to ask of any assessment you use

Whether or not you ever look at Evala, the shift is worth holding onto, because the cost of measuring the wrong thing lands on your floor either way. A few questions worth putting to any tool in your stack:

Does it separate the skills that fail independently, or average them into one number that hides the weak one? Does it test a real conversation that moves, or a single static prompt? Does it tell you anything about fit and aptitude, or only about language? Does it predict performance in the seat, or just confirm proficiency in the abstract?

The hiring question has moved on. Twenty years ago, "can they speak the language" was a reasonable proxy for "can they do the job." It is not anymore, and the operations that keep treating it that way will keep paying for the gap, one early exit at a time. The fix is not a harder language test. It is an assessment that finally measures what the job actually takes.

Evala began by scoring voice and written communication as separate signals and assessing how candidates handle real, multi-turn conversations. We are building toward a fuller readiness assessment that reads comprehension, aptitude, role fit, and domain readiness, so you can put the right person in the right seat. If you want to see where it stands today, book a demo.

Sources