Guides
Translation agency service profitability compared
Compare translation agency service lines by contribution per internal hour, and see which named lines earn most and why the top rate card entries rank last.
What to take away
- The highest rate per word is often the lowest earner, and translation agency service profitability follows contribution per internal hour, not the headline rate.
- Certified translation, transcreation, small rush jobs, one-off documents and multimedia carry hidden internal hours that never reach the invoice.
- Repeat volume, revision, terminology maintenance and software string updates look modest on the rate card and usually earn more per internal hour.
- Repetition compoundstranslation memory and termbases cut cost on the next project without cutting revenue, provided your agreement does not pass every match discount straight through.
- Run the ranking on your own last twenty projects. A ranking from an article is a ranking of somebody else's business.
The service lines, by name
Each line below carries the tool, standard or provider a general agency meets when it sells that line. Cost structure differs by line, and cost structure decides the ranking.
Service lines and pricing models
Service line
- Certified translation
- Per page plus fee
- Standard translation
- Per word, volume tiers
- Transcreation
- Per concept, hour, campaign
- Localization engineering
- Per hour or bundled
- Website localization
- Per word or retainer
- Subtitling and voice-over
- Per minute of media
- Interpreting
- Half-day or full-day rate
- Terminology work
- Project fee plus retainer
Pricing model
- Certified translation
- Standard translation
- Transcreation
- Localization engineering
- Website localization
- Subtitling and voice-over
- Interpreting
- Terminology work
| Service line | What it is | Typical pricing model |
|---|---|---|
| Certified translation for USCIS, courts and universities | A translator's signed certificate of accuracy attached to the text. US immigration accepts a statement from any competent translator; Germany and Spain appoint court-sworn translators instead. | Per page, typically $25 to $45 in the US market, plus a certification or notary fee |
| Standard document translation | Contracts, manuals and reports handled in a CAT tool such as SDL Trados Studio, memoQ or Phrase, where the match analysis sets the discount | Per word, typically $0.10 to $0.20 for common European language pairs, with volume tiers |
| Transcreation | Rewriting a headline or tagline for a target market rather than translating it. Wordbank and Freedman International sell it as a separate brief. | Per concept, per hour or per campaign |
| Localization engineering | File preparation and layout rebuilding in Adobe InDesign, Figma or markup, done before the linguist starts | Per hour, or bundled into the per-word rate |
| Website and software localization | Interface strings and help content pushed through Lokalise, Crowdin, Smartling or Transifex | Per word with match discounts, or a monthly retainer |
| Subtitling and voice-over | Timed subtitles, dubbing and narration. Streaming buyers such as Netflix publish timed-text style guides that vendors must follow. | Per finished minute, typically $8 to $25 |
| Interpreting | Spoken work on site, over the phone or by video. LanguageLine Solutions, Propio and Voyce sell it that way. | Half-day or full-day rate, plus travel |
| Terminology and glossary work | A client termbase in TBX format, maintained in SDL MultiTerm or memoQ | Project fee, then a small maintenance retainer |
| Revision and post-editing | Checking a client's own translation, or editing machine output to the ISO 18587 post-editing standard | Per hour or per word at a discount to translation |
That list is not the ranking. The ranking comes out of your own numbers.
The calculation
Take one completed project and pull four values from your records.
Calculating line earning rate
- Pull R, C, H and L from records
- Contribution = R - C
- Earning rate = (R - C) / H
- Sum R, C and H over a full quarter
- Line earning rate = (sum R - sum C) / sum H
- Rank lines by earning rate
- R
- = what the client was invoiced, excluding pass-through costs you did not mark up
- C
- = what you paid linguists, revisers and any external specialist
- H
- = internal hours your own people spent. Project management, file preparation and query handling come first. Checking, delivery and the invoice come after.
- L
- = the service line the project belongs to
Two numbers follow.
Contribution = R - C
Earning rate = (R - C) / H
The second number is the one that matters. Internal hours are the scarce resource, not word volume. An owner's week does not expand because a job pays well per word.
Run it across a line, not a project
One project tells you nothing. Do this for every project in a line over a full quarter, then sum:
Line earning rate = (sum of R - sum of C) / (sum of H)
Rank your lines by that figure. The order will not match your rate card, because the rate card prices words. It does not price the phone call to the consulate.
Which lines earn most, and why
Repeat volume for an established account usually ranks first. The translation memory and termbase already exist, so cost falls while the negotiated rate holds. A Lokalise or Phrase account pushing the same product strings every month is the clearest case.
Revision of a client's own translated text ranks high for the same reason. Those jobs are short, predictable and easy to schedule. The terminology work was done by somebody else.
Terminology maintenance earns well out of proportion to its fee. A TBX glossary that removes twenty queries a month across every other line for that client is the cheapest hour you will ever bill.
Software string updates on a live product sit close behind. The process is built, repetition is high, and the work arrives in batches you can schedule.
Certified translation usually ranks last despite the highest per-page rate. The internal hours explain it: confirming what USCIS or a German court accepts, printing, signing, booking a courier, then the call when a document is rejected. None of that reaches the invoice.
Transcreation ranks low for a different reason. Billed per concept or per hour, it still needs a brief, several options, a client debate about which one, and a second round. The debate is unbillable.
Small rush jobs are the quiet loss. Minimum charges look healthy per job, while setup, delivery and invoicing cost the same as a large job on a fraction of the revenue.
Multimedia and interpreting carry their own hidden hours: script freeze chasing, retakes, synchronization, travel and room setup. Bill them as separate lines. Never average them into written translation.
Three effects the arithmetic misses
Repetition compounds. Every project for the same client adds to the memory and the termbase. That lowers C on the next project without lowering R, provided your agreement does not automatically pass every match discount through. Decide that policy deliberately; it is one of the sharper questions in how translation pricing is built and where margin survives.
Concentration is a cost. A line earning well from a single client is a relationship, not a line. Price it accordingly and count the risk.
Some lines buy other lines. A small terminology engagement often converts into steady volume, so its own earning rate understates it. That is a sales argument, not an accounting one, and it belongs in how proposals and pricing conversations are structured.
What you need in place to measure this
You cannot compute H without internal hours recorded against a project, and you cannot compute C without subcontractor costs filed against that project. A timer such as Toggl Track or Harvest, tied to the job number, gives you H. This is a records problem before an analysis problem.
The IRS list of business records worth keeping sets the general expectation that records support the transactions they represent. A project-level cost file does double duty.
Two cautions on the result. A line that earns well for you says nothing about what another agency should sell.
Describing your ranking publicly as an industry fact is an advertising claim you would have to support. The FTC guidance on truthful advertising for small businesses covers what a claim has to be backed by.
Written translation and spoken interpreting also differ in shape, as the occupational description of the two professions sets out. Do not average them into one line and expect the result to mean anything.
Once you know which lines earn, staffing follows, and recruiting and training against the lines you actually sell becomes a narrower problem. The lines you lead with then shape how the agency presents itself and attracts the right buyers, and the menu itself is described in what each language service line involves.
Common questions
Which service is the most profitable in this industry?
Nobody can answer that for you, and any page that names one is guessing about your cost base. The calculation above is the answer, and it takes an afternoon with your own records.
Should I drop a line that earns poorly?
Not automatically. Ask first whether internal hours can fall. A line often earns badly because the process around it was never built, not because the work is unprofitable. Fix the process, measure again, then decide.
Does higher volume fix a low earning rate?
Only where internal hours do not scale with volume. If every job needs the same manual setup, more jobs means proportionally more of your scarce hours, and the rate stays where it was.
How often should I rerun this?
Once a quarter is enough to see movement, and after any change to your tooling or your review model, because both move internal hours directly.
