A junior reviewer once flagged my translation because I'd "forgotten" to convert psi to bar in a wellhead pressure spec. She was trying to help. She was also one keystroke away from feeding a rig crew a number that didn't match the gauge in front of them.
That's the thing nobody tells you about numbers in this work. The translator's instinct is to serve the reader — make it local, make it native, smooth every edge. In oil & gas that instinct is a liability. A number in a drilling program, a mud report, a pressure test certificate isn't prose. It's an instruction that has to agree with a physical instrument, a regulatory filing, and three other documents nobody showed you.
The unit is part of the noun
When a document says 3,000 psi, that's not "pressure, expressed in an American unit I should tidy up." It's the number the technician reads off the equipment, the number in the API spec the contract references, the number the client's HSE department signed off on. Convert it to 206.8 bar and you've created a value that exists nowhere else in the project. Now a Turkmen-speaking operator is holding a document that disagrees with his gauge by a rounding error, and he has to decide, at 2 a.m., which one to trust.
Same with barrels. A field reports production in bbl. Somebody wants it "clarified" into cubic meters, so 1,000 bbl quietly becomes 158.99 m³ — except the sales contract, the tax filing, and the daily report all still speak barrels. You've introduced a translation layer between two numbers that were supposed to be identical. Reconciliation nightmares are born exactly here.
My rule is blunt: I don't convert units in oil & gas unless the client tells me to, in writing, and gives me the target and the rounding convention. Default is preserve. If the source says 5,000 ft of casing, my Turkmen says 5,000 ft. I'll translate casing. I won't touch 5,000 ft.
Where it actually gets dangerous
The traps aren't the obvious conversions. They're the units that look the same and aren't.
Ton. English "ton" is a landmine. Short ton (2,000 lb, US), long ton (2,240 lb, UK), metric tonne (1,000 kg). A source document that just says "ton" and a translator who assumes metric can be off by ten percent on a cargo weight. I don't guess. I query. "Which ton?" is a boring email and a cheap one.
API gravity. People see a degree symbol and reach for Celsius. API gravity in °API is not temperature and does not convert to anything Celsius-shaped. Leave it. If you don't know what a unit is, that's the loudest signal in the world to stop translating and start asking.
Comma and period as decimal separators. Russian and Turkmen technical documents lean toward the comma decimal; a lot of the English source I get uses the period, and the thousands separators flip too. 1,500 in one convention is fifteen hundred; in another it's one-and-a-half. A single misread separator turns a flow rate off by a factor of a thousand. I lock this down per project and I do not let the CAT tool's autocorrect "help."
Gauge vs absolute pressure. psig and psia are not the same number and the letter that distinguishes them is exactly the letter tired eyes drop. If the source carries the distinction, I carry it. If it doesn't and the context matters, I flag it.
The part that's really a legal problem
Half of oil & gas translation is legal translation wearing a hard hat. The technical spec lives inside a contract, and the contract references standards — API, ISO, GOST, whatever the parties agreed on. Those references are not to be localized or paraphrased. API Spec 5CT stays API Spec 5CT. You don't translate the code, you don't "find the Turkmen equivalent standard," and you certainly don't convert the tolerances the standard defines into your preferred units to be helpful. The number in the contract is a legal commitment. Change its form and you've arguably changed the obligation.
This is where the "just make it read naturally" crowd gets people into arbitration. In marketing copy, natural wins. In a supply agreement for OCTG, the natural-sounding sentence that quietly rounded a wall thickness is the one that gets read aloud in a dispute.
So my deliverable in this domain often looks less finished than a good literary translation. Units untouched. Standard codes intact. Separators consistent and matching source intent. A translator's note where I found an ambiguity instead of a silent fix. It's not laziness. It's knowing that in a document meant to match a gauge, the most professional thing I can do to a number is nothing at all — and to say clearly, in the note, why I left it.
The reviewer who flagged my psi eventually agreed. The rig crew never knew there'd been a debate. Good. That's the job — the number matches the gauge, and nobody ever has to think about the person who made sure of it.