English
繁體中文
The call came from Stuttgart, three days after the German site went live. A dealer had configured a wardrobe that came out 78 wide, and the customer had measured her alcove at 80. She ordered. The wardrobe was built in inches. Somewhere in a build that had passed every QA checklist, nobody had localized the number itself, and a multilingual product configurator turned out to be a harder thing than the seven translated interface files suggested.
That launch is a composite of several we have seen, and the sequence below is roughly how it went, in the order the problems surfaced. Each one was cheap to fix afterwards and expensive to have missed.
Everything looked translated on day one
The interface was in German. Buttons, tooltips, error messages, the confirmation email. The vendor's QA sheet had a column for language and it read complete.
What the sheet did not have columns for: whether the numbers meant the same thing to a German buyer, whether the fabric names were words anyone in that market would use, and whether a price shown as a total was the price a German customer is legally allowed to see. All three were wrong. All three were invisible from the office where the build was tested.
The tape measure conversation
Furniture is bought with a tape measure in hand, which makes dimension display the first thing to get right and the first thing to get wrong.
Our model was authored in millimetres, because that is how the factory works. The US site displayed inches, converted at the display layer. The German build inherited the American display setting, so the shopper read a number that was plausible as centimetres and was actually inches. Seventy-eight of them, in a market where 78 means something a third smaller.
Conversion is the easy part. The roundings are where judgement lives: 198.12 centimetres is arithmetic, not a dimension any buyer wants to read, so each market needs a display precision chosen by someone who knows how the product is discussed locally. Tolerances differ too, and a tolerance that reads reassuring in one country can read alarming in another. None of it survives if the model's own accuracy is shaky to begin with, which is the groundwork covered in furniture 3D viewer accuracy.
The fabric that had three names
A week later, a dealer asked for a fabric by a name that did not exist in our system. The German term she used was normal in her market, absent from our catalogue, and the machine translation of our English name meant something slightly different, so she had been quoting the wrong finish for a week without noticing.
Materials are where dictionaries fail hardest, and the place a 3d configurator localization project loses the most time without noticing. A weave gets named for its construction in one country, by a brand convention in another, and by a generic local word in a third. What you need is a table with three columns per market: the local commercial name, the internal code, and the phrase a customer would type into a search box. The code travels to the factory. The other two travel to the screen and the page title. They are rarely the same string, and treating them as one is how a plausible word ends up describing the wrong product.
The price that was legally wrong
Currency conversion was working perfectly, which is why this one took a while to spot.
In that market the displayed price has to include tax. Ours excluded it, because the base build was American and the American store had always shown tax at checkout. A dealer presenting a total to a customer was presenting a number the customer would not pay. There was rounding, and one more question nobody had considered until finance asked it: what happens to a configuration saved in March and reopened in June, when the price list has moved.
A configurator shows a total. A PDF catalogue never had to, which is why the pricing decisions that used to live quietly in the checkout now need an owner on the 3D side. Deferring them does not break the launch. It breaks the first international order, usually in a meeting where somebody is defending a number.
Where the layout broke
German compound words pushed two buttons off the panel. That was the visible one. The invisible one would have been Arabic, where the panel reverses direction while the 3D canvas keeps its own coordinate system, and where an interface that reads gracefully in English starts fighting the user for orientation.
The habit that catches both: build the control panel around the longest plausible string instead of the English one, and run the full flow in the two most difficult languages you intend to support before anyone signs off. Failures cluster in the same three places, option labels, price totals and error text, so the test stays short if you know where to look.
What search engines saw
Six weeks in, the German pages had traffic from paid ads and almost nothing organic. The cause was architectural. The localized experience lived on the same URL as the English one, behind a language switcher, which gives a crawler one page in one language and no way to index the others.
Each market needs its own URL, a title and description written by a person who sells in that language, and hreflang pointing at genuine equivalents rather than at a redirect chain. Product and material names matter more than usual here, because the local commercial name is what people search, and it is almost never the translated version of your English name. Performance budgets deserve a second look per market as well; the constraints in web 3D configurator performance are not constants, and a market on slower mobile networks feels every megabyte differently.
The part nobody budgeted
Two months after fixing all of the above, a dealer emailed in German asking about a lead time. The inbox was monitored by someone who read English.
A configurator that speaks German and answers questions only in English gives back about half of what it earned. The arrangement that works is narrow and unglamorous: one native speaker per market who can read a configuration payload, plus a written escalation route for what they cannot resolve. Put it in the launch plan beside the translation budget, and if dealers across several markets are involved, the sequencing in rolling out a 3D configurator to a dealer network matters more across borders than most teams expect.
The multilingual product configurator lesson, after several launches, is that translation is the smallest item on the list and the only one everyone remembers to do. Units, material naming, price presentation, layout behaviour, per-language URLs and after-sales language decide whether a market feels served or feels like an export channel. If a second market is on your roadmap and the first is already live, tell us where it is going and we will walk the list with you before a dealer finds the gap with a tape measure.
FAQ
Q: Can I just translate the interface strings?
Translation covers labels and leaves the hard parts untouched. Units, material naming, price presentation and layout account for nearly every problem we have seen, and none of them is solved by a translation file.
Q: Should each market get its own configurator build?
Almost never. One build with per-market settings for units, naming, tax display and copy is cheaper to maintain and stays consistent, provided the asset library is shared and the display layer is properly separated from the model.
Q: How do I know a market is actually finished?
When someone who speaks that language has placed an order end to end and the invoice matched what the screen showed. That single test finds more than the rest of the checklist combined.
