On this page
Five cost categories you can actually measureA four-week methodEstimate it: a simple calculatorWhat changes with AI translationPresenting the resultFAQArticles about translation risk love big numbers: billions lost, percentages of customers who leave. Most are hard to trace to a source, and none of them describes your company. When a localization manager takes such a figure to a finance director, the conversation usually ends with "that's not us". A better approach is to measure what poor translation costs your own business, with your own data. The numbers will be smaller and far more persuasive.
Five cost categories you can actually measure
| Category | What it includes | Where the data is |
|---|---|---|
| Rework | Staff hours spent fixing translations; re-translation fees; repeated review rounds | Time tracking, vendor invoices, project management tools |
| Support load | Tickets caused by unclear or wrong translated content | Support system, tagged by language and cause |
| Lost conversion | Lower conversion on translated pages than comparable pages in the source language | Web analytics, A/B tests, checkout funnels by locale |
| Returns and refunds | Returns where the customer cites misleading or unclear information | Order system, return reason codes by market |
| Delay | Launches, filings or contracts held up by translation problems | Release records, legal and filing calendars |
A sixth category, legal and reputational exposure, is real but rarely measurable in advance. Keep it separate, as a list of risks with examples, rather than putting a speculative figure into the same table.
The few figures worth quoting have a method behind them. CSA Research's 2020 survey of 8,709 consumers in 29 countries found that 76% prefer to buy products with information in their native language. That measures preference, not your losses, which is why the categories below are about your own data.
A four-week method
- Pick one market and one content type. For example, the Korean help centre, or Korean product pages. Small and measurable beats broad and vague.
- Tag the data. Ask support agents to tag tickets where the cause is translated content. Add a reason code for returns that cite product information. Ask reviewers and in-country staff to log time spent fixing translations.
- Compare with a baseline. Compare conversion, ticket rate and return rate for the translated content with the source-language equivalent, or with the same content before and after an improvement.
- Price it. Multiply hours by loaded hourly cost, tickets by average handling cost, lost conversions by average order margin.
- Annualize carefully. Scale four weeks to a year only where the volume is steady, and show the assumption.
- Compare with the fix. Put the measured cost next to the cost of better translation for that content: post-editing, review, a glossary, a voice guide.
Estimate it: a simple calculator
Enter your own figures for one market and one content type over a month. The calculator only multiplies what you give it; it contains no industry averages.
Example values are placeholders, not benchmarks. Replace them with your own. Lost conversion is left out on purpose: measure it separately with an A/B test or a before-and-after comparison.
What changes with AI translation
Machine translation lowers the visible cost of translation and moves cost elsewhere. Translation invoices go down; rework, support tickets and review time can go up if output is published without adequate review. Measuring the categories above, before and after a switch to machine translation, is the only reliable way to know whether the change saved money. Often the answer differs by content type: a saving on help articles, a loss on product pages. That is an argument for routing content by risk; see when to use machine translation and when to use people.
Presenting the result
Keep it to one page: the content type and market, the measured costs by category, the assumptions, the proposed fix and its cost, and the risks you could not quantify. Finance teams respond better to "we measured 12 hours a week of rework on Korean product pages" than to any figure about the global economy. For a broader view of where costs hide, see the real cost of choosing the cheapest patent translation service and the real cost of cheap Korean translation; for keeping quality affordable, read translation QA methods that reduce cost without hurting quality.
FAQ
Why not use published statistics about the cost of bad translation?
Many are hard to trace, and none describes your business. Your own data is more credible.
What is the easiest cost to measure first?
Rework: staff hours and re-translation fees.
How do we put a number on reputational damage?
Usually you cannot. Record it separately as a qualitative risk.
Measuring your translation costs
We help set up the tracking and review the results for your Korean content. Talk to us →