A European pharmaceutical company ran their entire localization pipeline through MTPE. Everything — clinical trial protocols, patient-facing materials, internal training documents, regulatory submissions. They chose Full MTPE for everything because Full MTPE “should be good enough,” and it cost 40% less than human translation from scratch. During a regulatory audit in Germany, the BfArM flagged their translated investigator’s brochure for semantic accuracy issues in three sections: dosage escalation tables, contraindication language, and adverse event reporting terminology. The MT had misinterpreted “titrate to tolerance” as a general instruction rather than a dose-limiting protocol, and the post-editor had missed it because the sentence read fluently. The company received a conditional approval with a mandatory correction within 30 days, a 12-week enrollment delay, and an estimated €1.8M in lost pipeline value. They saved €340K on translation. They lost €1.8M on the drug program.
This isn’t an argument against MTPE. It’s an argument for knowing what kind of content you’re putting through MTPE and choosing the quality tier accordingly. Most organizations don’t. They either default to raw MT for everything to save cost, or Full MTPE for everything because they’re risk-averse, or they apply MTPE inconsistently because nobody has done the content profiling work that would tell them where each tier is appropriate.
MTPE has two established quality tiers, and they are not interchangeable. Light MTPE is designed for content where the primary requirement is comprehensibility. The post-editor checks for major errors — mistranslations, omissions, significant grammar problems — and makes corrections where meaning is distorted. They do not rewrite for style, terminology consistency, or register. Light MTPE assumes that the MT output is mostly correct in meaning and needs targeted fixes, not a rewrite. The target quality is “understandable and not misleading.” Full MTPE is designed for content where the output needs to be indistinguishable from human translation. The post-editor corrects meaning errors, but also addresses terminology consistency, stylistic appropriateness, register, tone, and fluency. They may substantially rewrite sentences. Full MTPE assumes that MT output has systemic issues beyond isolated errors and requires comprehensive intervention. The target quality is “publication-ready.”
Here’s where most organizations go wrong: they assume Full MTPE is a universal solution that makes MT output safe for any content type. It isn’t. Full MTPE improves MT output. It does not make MT output equivalent to human translation performed by a domain specialist. A post-editor who is not a pharmaceutical regulatory expert will not catch a misinterpretation of “titrate to tolerance” regardless of how many passes they do, because the sentence reads fluently and the error is in the domain semantics, not the language. The quality of MTPE is bounded by the domain expertise of the post-editor.
Content profiling is how you decide. The matrix has four dimensions: visibility, regulatory exposure, linguistic complexity, and shelf life. I’ll break each one down.
Visibility: who reads this content and what happens if it’s wrong? Internal knowledge base articles read by your own team? Low visibility, low consequence of minor errors — Light MTPE is appropriate. Customer-facing product documentation? Medium visibility, errors damage credibility — Full MTPE. Marketing copy, legal contracts, regulatory submissions? High visibility, errors carry financial and legal consequences — human translation from scratch, with MTPE used only for reference drafts.
Regulatory exposure: does this content go to a regulator, and if so, what happens if it’s wrong? Clinical trial materials, drug labeling, medical device IFUs — these go to agencies like the EMA, FDA, BfArM, PMDA. Translation errors can delay approvals, trigger audits, or cause patient safety issues. Full MTPE is not sufficient for this category. Human translation by domain specialists with regulatory review is the minimum acceptable standard, and MTPE should be used only for drafting, not for the final version. Internal SOPs, training materials for non-critical processes, meeting minutes — these have low regulatory exposure and Light MTPE is fine.
Linguistic complexity: does this content use specialized terminology, ambiguous constructions, or culturally specific references? Legal contracts (consideration, indemnification, force majeure) have high linguistic complexity because terms carry legal meanings that MT systems frequently misinterpret. Marketing copy has high linguistic complexity because it relies on connotation, cultural reference, and emotional resonance that MT handles poorly. Technical specifications for industrial equipment have moderate complexity — the terminology is specialized but the sentence structures are usually straightforward. User interface strings and product descriptions for e-commerce have low complexity and are good candidates for Light MTPE.
Shelf life: how long will this content be used, and how often does it change? A quarterly earnings report has a shelf life of months and changes every quarter — Light MTPE with targeted human review of key financial terms is often sufficient. A software license agreement has a shelf life of years and may be litigated — human translation with legal review. A social media post has a shelf life of hours — raw MT or Light MTPE.
The budget allocation that works: profile your content across these four dimensions, assign each content type to a quality tier, and apply that tier consistently. A mid-size SaaS company localizing into 8 languages typically finds that about 40% of their content volume is suitable for Light MTPE (internal docs, UI strings, knowledge base), 35% requires Full MTPE (product documentation, customer communications, onboarding materials), and 25% needs human translation from scratch (legal, marketing, regulatory). That allocation produces roughly 50-55% cost savings versus all-human translation while keeping risk where it belongs. The pharma company that ran everything through Full MTPE and lost €1.8M? Their content profile would have put the investigator’s brochure in the 25% human translation category. They would have saved less overall but avoided the audit finding entirely.
The dirty secret of MTPE budget optimization is that it’s not about finding the cheapest quality tier for each content type. It’s about finding the tier that produces acceptable quality for that content type, and the definition of “acceptable” changes depending on who reads the content, what happens when it’s wrong, and how long it needs to be right. Content profiling answers those questions. Everything else is guessing.
Artlangs Translation provides MTPE translation services across 230+ language pairs: content profiling and quality tier assignment, Light MTPE for high-volume low-exposure content, Full MTPE with domain-specialist post-editors for publication-ready output, and human translation with legal and regulatory review for high-risk content. Because the right MTPE tier saves money. The wrong one costs more than human translation ever would.
