Two clients pay me the same rate. One I'd take at half the price. The other I'd drop tomorrow if I had something to replace it with. Same cents per word, wildly different jobs. And the only way I know that is because I stopped trusting the per-word number years ago.
Here's the thing nobody writes on their rate sheet: the word is a lousy unit of measurement. It measures the text. It doesn't measure the work. Two 3,000-word files can cost me two hours or two days depending on the state they arrive in, the number of queries buried inside them, and how many rounds of "can you just check row 47 again" come after delivery. The per-word rate flatters the messy job and punishes the clean one. Same money, but one earned me €60 an hour and the other earned me €18.
What the per-word rate hides
Run the tape on a real job. Client A sends a clean, bilingual XLIFF, a proper term base, a reference PDF that actually matches the source, and a PM who batches her questions into one email. I translate, I review, I deliver. Done. Client B sends a PDF someone scanned crooked, no glossary, half the strings marked "context: N/A," and then three separate follow-ups over the next week because marketing changed the headline and legal changed a clause and someone in Almaty wants the Cyrillic version too, which was never in the PO.
Both invoices say the same thing. My bank account eventually agrees. My calendar does not.
The ATA compensation survey, the one everyone quotes, is built on per-word and per-hour figures. Useful as a sanity check, sure. But it can't tell you the part that actually decides whether freelancing pays — how many unpaid hours are stuck to each paid word. There's a study out this year on freelance pricing in Belgium and Switzerland, looking at how translators set rates and whether they're satisfied with what they earn. The interesting bit isn't the rates. It's the gap between rate and satisfaction. People charging decent numbers who still feel underpaid. That gap is the unbilled hour. It's the query email you answered at 9pm. It's the invoice you chased for the third time.
Just track it. Badly is fine.
I keep a spreadsheet. Nothing fancy. When I start a job I note the time. When I stop, I note it. Every query email, every reformatting scramble, every "quick" extra file — it goes on the row for that client. At the end of the month I divide what each client paid by the hours they actually cost me. That's the effective hourly rate. That's the only rate I care about.
The results are rarely what the per-word rate predicts. My best client on paper — highest cents per word — sits mid-table on effective hourly, because their source files are always a mess and their reviewers relitigate every choice. A mid-rate agency that sends locked, pre-processed files with a clean TM and never touches me after delivery? Top of the table. I'd give them a discount before I'd give it to the "premium" one, and I'd mean it.
You don't need to track forever. A month, honestly done, tells you most of what you need. Two months and you can rank the whole roster.
Why this matters more now, not less
The volume is thinning. The Washington Post ran a piece this year on translators feeling AI eat their work, and it's not paranoia — the raw wordcount that used to flow to freelancers is getting routed through machines first. Slator's year-in-review on how AI reshaped the industry says the quiet part: the easy, clean, high-volume work is the first to go. What's left for humans skews harder, messier, higher-stakes. More queries, more judgment, more of exactly the stuff that doesn't fit neatly into a per-word cell.
Which means the gap between per-word rate and effective hourly rate is going to widen, not close. If your remaining work is the difficult residue — the legal clause that can't be wrong, the marketing line that has to actually land in Turkmen, the oil-and-gas spec where a comma shuts down a rig — then measuring yourself in words is measuring the wrong thing entirely. You're not selling words anymore. You were never really selling words. You were selling the hour, and the judgment inside it.
So when a new agency onboards you and asks for your per-word rate, give them one. Fine. But keep your own books in hours. When they push for a discount, you'll know exactly which of your current clients you can afford to lose and which you'd fight to keep — and it won't be the ones with the fattest per-word number.
The client who respects your time is worth more than the client who pays for your words. The spreadsheet just makes it impossible to pretend otherwise.