The Skill Gap Nobody Tests For 

What 122,000 English assessments reveal about the one skill hiring managers never check 

Picture the candidate everyone wants: C1 certified, reads a technical spec without blinking, writes emails that need no editing. Then the first client call happens, and something doesn’t quite add up. They ask the client to repeat themselves twice. They answer the question that was asked three sentences ago. On paper, this person’s English is excellent. On a call, it isn’t quite working. 

That gap shows up often enough to be worth asking whether it’s really a communication problem or a measurement problem. A large enough dataset can answer that, and the answer turns out to be the second one. 

A pattern invisible in one test, obvious in 122,000 

We looked at more than 122,000 completed English assessments, each broken into four skills: grammar, vocabulary, reading and listening. Three of the four cluster closely together, with grammar and reading averaging 73% and vocabulary at 70%. The fourth doesn’t. Listening averages 64%, a full nine points below the strongest skill, and the gap holds steady across the whole population rather than a handful of outliers. 

The figures come from an online CEFR assessment taken by candidates in more than 190 countries, one that scores grammar, vocabulary, reading and listening independently instead of folding them into a single number. That separation is the only reason a gap like this becomes visible at all: compress four skills into one score, and a weak listening result simply disappears into a decent average. 

On an individual test, a low listening score reads as noise. Somebody had a bad connection, a noisy room, an off day. Across 122,000 sessions, it stops being noise and becomes the single most consistent weak point in the dataset. 

Why it’s the skill interviews never catch 

Grammar and vocabulary show up everywhere a hiring process already looks: the CV, a writing sample, a short scripted chat. Listening comprehension in real time, fast, informal, interrupted, accented, rarely gets tested at all, because a standard interview has no obvious moment built to expose it. A candidate who has prepared answers to the five questions they expected can sound confident without ever proving they can follow a conversation nobody scripted for them. 

Self-assessment doesn’t fix this either. Most learners rate their own listening ability higher than their tested result, for much the same reason people overrate how well they follow a podcast at double speed: comprehension feels complete in the moment, and the gaps only show up when something gets missed and nobody points it out. 

Why it matters more with every year of remote hiring 

The stakes are higher than they used to be. A misread email gets caught by a second read. A misheard instruction on a live call, in a language that isn’t the listener’s first, often doesn’t get caught until the deliverable comes back wrong. Distributed teams run on close to nothing but the skill this dataset shows is weakest: real-time listening, under time pressure, without the visual and contextual cues a face-to-face conversation provides for free. 

What to actually check before you hire 

Ask for a skill breakdown, not a single score. A B2 or C1 label is an average. It doesn’t tell you whether that average is built on strong reading and weak listening, or the other way round. When a result arrives as a skill-by-skill certificate, you can see in seconds where the weak point sits and weight the interview accordingly, rather than finding out on a client call three months in. 

Build one unscripted moment into every interview. Ask a follow-up question that wasn’t on the list, or have the candidate summarise something you just said in their own words. Prepared answers don’t test listening. Unprepared ones do. 

Don’t assume a high level cancels out a weak skill. A C1 candidate with below-average listening can still outperform a B2 candidate overall, but the specific risk, missed instructions on a live call, doesn’t disappear just because the headline label is impressive. Match the check to the risk, not just to the score. 

A gap that’s cheap to check 

None of this is a dramatic failure of English teaching or of hiring practice. It’s a blind spot created by the fact that reading and vocabulary are easy to observe and real-time listening isn’t. At the scale of over a hundred thousand tests, that blind spot has a number attached to it: nine points, consistent, and worth asking about before the client call rather than after it. 

International English Test (internationalenglishtest.com), an associate member of ALTE (the Association of Language Testers in Europe). Figures are aggregate and anonymised, drawn from completed assessments on the platform.