Call QA in Catalan, Basque, and Galician: Closing the Regional-Language Blind Spot

Walk the floor of a contact center in Barcelona, Bilbao, or Vigo and listen for five minutes. You will not hear one language. A single agent might greet a customer in Spanish, switch to Catalan when the customer does, and take the next call in Spanish again. In Galicia, customer conversations slide between Galician and Spanish mid-sentence. In the Basque Country, some customers simply expect service in Euskara — and choosing to serve them in it is, for many brands, both a service standard and in some contexts a regulatory expectation.

Now ask a harder question: what happens to those calls when they reach your QA process?

In most operations, the honest answer is: nothing good. They're skipped, mangled, or scored wrong. This post is about that blind spot — why generic speech analytics tools create it, what it costs, and how we built Caller.ee to close it.

The Multilingual Reality of Spanish Contact Centers

Spain's co-official languages aren't a niche. Catalan is spoken by millions of people across Catalonia, Valencia, and the Balearic Islands; Galician and Basque each anchor daily life in their regions. Public administrations, utilities, banks, insurers, and telcos operating in these territories field regional-language calls every single day — and many are formally committed to offering service in them.

BPOs feel this most sharply. Win a contract with a Catalan utility or a Galician public service, and multilingual handling isn't a nice-to-have in the SLA — it's the job. Agents are hired for these language skills, paid for them, and trusted with the customers who use them.

Trusted — but rarely verified. Because when it's time to evaluate quality, most tooling quietly assumes the whole world speaks English, or at best "mainstream" Spanish.

What Generic Tools Do With a Catalan Call

Speech analytics vendors typically train and tune for English first, add the large global languages, and stop. Feed a Catalan, Basque, or Galician call into an English-first pipeline and one of three things happens:

• It's mistranscribed as Spanish. Catalan and Galician are close enough to Spanish that an automatic language guesser often picks Spanish and produces confident nonsense — words that look plausible and mean something else, or nothing. Basque, which isn't related to Spanish at all, fares even worse.
• It's transcribed as gibberish. The transcript comes back as word salad, and any AI analysis built on it is analysis of noise. Scores from a garbage transcript aren't low-quality data — they're fiction with a percentage sign.
• It's silently excluded. The team learns the tool "doesn't do those calls," and regional-language calls drop out of the QA universe entirely.

All three outcomes converge on the same place: the calls your operation handles in Catalan, Basque, or Galician are invisible to quality management.

The Costs of the Blind Spot

Unfair — and legally awkward — agent evaluation

If QA only functions on Spanish and English calls, your bilingual agents are evaluated on a fraction of their work — and often on the fraction they're less specialized in. An agent hired specifically for excellent Euskara service gets scored only on her Spanish calls. Her strongest skill is professionally invisible. Over time that skews coaching, reviews, and bonuses in ways that are hard to defend if anyone ever asks you to.

Compliance risk doesn't switch languages

Mandatory disclosures, consent language, contract conditions read aloud, retention offers with regulated wording — these obligations apply to a call in Catalan exactly as they do in Spanish. If your monitoring can't understand the call, you cannot demonstrate the disclosure happened. For regulated industries and public-sector contracts, "we don't QA calls in that language" is not a sentence you want to say in an audit.

Inconsistent quality where it's most visible

Customers who choose to speak Catalan or Galician are often the customers most attentive to being served well in it. Leaving exactly those interactions unmonitored means your quality program is weakest precisely where customer expectations — and sometimes contractual commitments — are highest.

How Caller.ee Handles Regional Languages

We made regional Spanish languages a first-class design goal, not a footnote. Caller.ee's transcription engine — running, like our whole pipeline, on our own European infrastructure — supports roughly 40 languages, and Catalan, Galician, and Basque sit alongside Spanish, English, Portuguese, French, German, Italian, and Arabic as fully supported options.

The mechanism matters: when you upload calls, you can explicitly set the language per upload. That one choice removes the failure mode that breaks most tools — the automatic guesser deciding your Catalan call is Spanish. Tell the pipeline the batch is Catalan, and it transcribes it as Catalan, with accuracy that holds up for real QA work rather than a rough gist. The same speaker separation applies as everywhere else on the platform: agent and customer identified even on mono recordings.

From there, everything downstream just works, because the rest of the pipeline operates on the transcript:

• Your scorecards apply unchanged. An Analysis Template's sections are natural-language prompts; the analysis evaluates the Catalan or Basque transcript against the same criteria — greeting, identification, disclosure, objection handling, next step — that you use for Spanish calls. One quality standard, every language.
• Knowledge-base grounding still holds. Templates grounded on your uploaded scripts and policies check the agent against your actual playbook regardless of the language the conversation happened in.
• Search covers everything. Transcripts are semantically searchable across your whole call base, so regional-language calls show up in investigations instead of being a separate, unsearchable pile.

What This Unlocks for Multilingual Operations

Fair evaluation across languages. Every agent is scored on all their work. Your Basque-language specialist finally gets credit — and coaching — on the calls she was hired for. Per-agent analytics reflect reality instead of a language-filtered sample.

One scorecard, consistent everywhere. COOs of multilingual operations can run the same template across a campaign's Spanish, Catalan, and Galician calls and compare like with like: same criteria, same 0–100% scale, same fail threshold. Quality reporting to clients stops needing an asterisk that says "except regional-language volume."

Compliance coverage without exceptions. Disclosure and script-adherence checks run on 100% of analyzed calls, whatever the language — which converts "we believe our Catalan calls are compliant" into evidence you can show.

A sharper pitch for BPOs. If you're bidding for work from brands with regional-language commitments, being able to demonstrate QA coverage in Catalan, Basque, or Galician is a differentiator most competitors can't match — because their tooling can't do it.

Hear It on Your Own Calls

If your operation handles calls in Catalan, Basque, Galician — or any of the roughly 40 languages we support — the test that matters is your own audio. Caller.ee is in public beta with a live free tier: 45 transcription minutes, 30 calls, and 15 analysis runs per month, free. Create an account at https://app.caller.ee, upload a few regional-language recordings with the language set explicitly, and run a scorecard against them. If the pilot needs more room, email info@sumgrey.com and we'll raise your limits.