Language translation for ecommerce is not primarily a technical issue, it’s a business issue. Translation costs money so any decision to translate has to be done within the context of Return On Investment (ROI). The cost is not just for the translation but also for the page management and its associated marketing to different countries or regions.This article is limited to just translation of written text. It’s also worth pointing out the difference between translation and localisation. Switzerland, for example, requires both French and German. Finally, it’s worth noting other types of translation that are not covered here include:

1. Translation of pricing e.g. Pounds to Euros.
2. Translation of images. For example, making use of ethnically relevant images for fashion model photoshoots  {ref}.
3. Translation of measurement. For example, UK foot size to Japanese foot size or metric to imperial {ref}.
4. Translation of layout. For example, different usability for right to left languages (e.g. arabic, hebrew, urdu, etc) {ref, ref}.

For a channel ecommerce business {ref} that is selling other company's brands, the calculations around page translation are reasonably straightforward. Calculations are based on two things:

1. Costs for translation. Translated pages may carry a higher conversion rate than ‘foreign’ language pages. Some references claim this to be a major factor {ref}, others suggest it to be a minor factor {ref}.
2. Costs for translated content deployment. For example, if you have your website available in four translated languages then you have four times as many pages to maintain, to manage and to market {ref}.

There are various levels of translation that might be considered for a website:

• Level 1; Leave it to Google.
• Level 2; Just translate a few key pages (e.g. legal terms and conditions).
• Level 3; Use automated translation on the whole website.
• Level 4; Use human translation on the whole website

To help appreciate the costs, if you have 10,000 products that you want translated into 5 different languages then you have 50,000 product pages to maintain and deploy. Next consider the time and cost for translating a product page. To do this manually could take a total of half an hour per product page. That means 5,000 hours per language. A full-time manual translator might be able to work hard for 5 hours a day for 260 days in a year {ref}. Using the following formula:

We get a total of 19.23 years of manual translation.If the work is spread across a team of four people, you still end up with four people working for around five years to complete the job.Now let’s consider the average product page lifetime. If you work in the fashion industry, then one might reasonably expect that at least half the catalog changes with the four seasons. In this situation, then if you had three months notice to prepare for the new season then you would need a minimum of 35 full time translators to ensure you had your translated web content ready for the new seasons launch.As of August 15th 2017, Amazon had 536,641,219 product pages. If Amazon chose to manually translate these into their ten most popular languages, then using the above formula I estimate that if Mr Bezos required that the products be available for sale in one month, he would need to employ an extra 25 million new staff ((536,641,219*0.5*10)/(5*(260/12))=24,768,056) - i.e. around five times the entire population of Scotland.Because of the costs associated with translation, most businesses just use their English website to sell to Europe as the majority of people in Northern Europe speak English. The following map from Wikipedia  {ref} shows the proportion of people in each European country that speak English:

The following is a list of the top 15 European countries by English speakers (in millions) {ref}:UK (59.6m)Germany (46.3m)France (23m)Italy (17m)Netherlands (15m)Poland (14.3m)Turkey (12m)Spain (10.4m)Sweden (8.2m)Belgium (6.3m)Austria (6.2m)Greece (5.5m)Denmark (4.8m)Switzerland (4.7m)Norway (4.5m)The list shows that there are potentially 178.8 million customers available across continental Europe that will be able to read and understand the website without translation. That said, conversion rates would always be expected to improve when content is translated into the native language.The following map is an alternative source highlighting the widespread use of English as a second language:

Top Level Domains (TLDs) v subfolders

The reason why we recommend clients use subfolders (e.g. /de/) over Top Level Domains (e.g. .de) is mainly due to SEO and much lower set-up/ongoing maintenance costs e.g.

• Significantly better SEO performance as all external links lift all pages across all sites;
• Easier setup of hreflang tags;
• Quicker to setup which means lower costs;
• Lower ongoing maintenance costs;
• Subfolders can be targeted via Google Search Console and Bing Webmasters to specific countries;
• Depending on your choice of Content Delivery Network (CDN), a geographically localised and targeted site is likely to use a local server - ie. local sites are likely to run faster.
• No need to buy individual SSL certificates for each hostname/domain;
• No need to pay for additional licenses for Magento extensions;
• No need to pay for multiple CDNs for each hostname/domain;
• No need to pay for multiple monitoring accounts e.g. Pingdom, Deepcrawl, Moz etc;
• No need for setup and manage Google search console for multiple domains;
• Everything tracked via one Google Analytics account;

A detailed analysis on this can be found here.Ecommerce platforms such as Magento use this approach by default. It’s worth noting that this is not possible using platforms such as Shopify. For these smaller, simpler platforms, the only approach is to use a different domain for each country i.e. website.fr, website.de etc {ref}.One issue that’s not often appreciated is that if you base your translations around the use of translation tools such as Google translate {ref} then a direct translation will be classed as duplicate content and, like all duplicate content, will suffer accordingly {ref}. You'll not be directly "penalised" i.e. only duplicate content pages will be hidden in search results in that language with other "non-duplicate" pages not affected.Magento has extensions that assist with translation. Two example services are Lingo24 and Gengo. Each offers different levels of translation quality associated with different levels of cost. It should always be noted that the quality of translation is invariably qualitative. Humans, like machines, are never perfect at translation. It’s also worth noting that services such as Google translate are getting better and better all the time.To reiterate the initial discussion, website translation is expensive and time consuming and should only ever be considered following an extensive ROI or cost-benefits analysis. The only language translation area not covered directly by the ROI calculation is the consumer perception of the brand {ref}. Ultimately, this is a judgement about how International a brand wants to be perceived. For example, Adidas is increasingly wanting to be seen as a global brand (rather than a German brand). Primarily, and for this reason, the website is now available in a wide range of languages e.g. here or here.The other thing that I always like to mention is demand factors. Why are customers buying from the UK? The number one factor for UK online exports is that consumers cannot buy the product in their own country (ref). This suggests they have already made the decision to purchase an international product, and as already shown, language is not a barrier for the majority of people. The language “barrier” is based on far more than understanding (Google translate can help you understand the content). If the consumer has already made the decision to buy from an overseas company, then this barrier is significantly reduced, and other areas (such as speed), play a much larger role.It is also worth mentioning that there are always companies that can benefit massively from translation, but it is the exception and NOT the rule. A lot of the time it depends on how differentiated the product offering is. If it is unique (e.g. Scottish Whisky) it is less important, if it is homogeneous (e.g. unbranded garden furniture) companies are competing directly with local suppliers (this is much harder to compete, and translation is often more relevant).Emil Stickland from Thrive Digital provided useful contributions to this article.