Home / Glossary / English as a filter
Communication
English as a filter
English as a filter is the hiring rule that a candidate reaches the client only if their written and spoken English would be sent to a customer without a rewrite. It is a screen on the person. It is not a claim about.
Also called customer-facing English screen, written English bar
The Philippines scores in the high band on the EF English Proficiency Index and sits near the top of Asia on that table. English is also an official language and the working language of the IT-BPM industry. None of that tells you whether the person in the 2:00 p.m. interview can write a refund email your buyer will accept. National indices measure test-takers. You are hiring one inbox.
How it works
Put the bar in the brief: customer-facing writing, not “conversational.” Then collect evidence before the founder’s calendar is used.
A 200-word sample on a real task beats a CEFR badge. “A customer paid twice. Write the email. Do not use a template.” A 15-minute live session on a messy thread beats a self-rating of “fluent.” Score three things separately: grammar that a customer will notice, tone that matches your company, and the habit of answering the actual question. People fail the third one more than the first.
Accent is not the filter unless the seat is voice and the buyer has a stated bar. Register is the filter for most 701am seats: inbox, care, books, research. A person can have a audible Philippine English accent and still write cleaner than a U.S. intern. The reverse is also common.
Example: 80 applicants. 30 pass a short written screen. 12 survive a live rewrite of a real customer thread. Three meet the founder. That drop is the filter working. A process that ships 15 “fluent” resumes to the client has not filtered. It has forwarded the country’s average.
How it differs
Accent and register are two different problems. Customer-facing English is the output standard. An English writing sample is the artifact. Taglish in an internal Slack is not a fail if the customer never sees it. Taglish in the customer thread is a fail. EF EPI is background. It does not belong in a candidate scorecard.
Common errors
Treating “from the Philippines” as the screen. Treating a call-center year as proof they can write as an EA. Ghostwriting the sample. If the prose arriving in month two does not match the sample, the filter was on the wrong author.