Author: Chris Jones, Independent Multichannel Retail Consultant
In a previous article I took a look at the effectiveness of comparing SEM costs from one country to the next, as part of the budgeting/planning process of considering whether to localise your e-commerce offer for a particular new country.
In that article, I explored using global brand-names as a possible methodology for doing so, as a way of eliminating language bias from the comparisons. The results were fairly reasonable, but the suspicion remains that, even if language was eliminated as an issue, cultural bias – the relative popularity of brands around the globe – still skewed the results.
In this follow up article, I take a look at an alternative method, one that is probably closer to how most retailers might actually tackle the issue in reality, which uses generic terms instead of brand terms.
There are a small number of retailers whose proposition is, deliberately, very consistent across the multiple geographies in which they operate. IKEA is one such. However an exploration of their international sites, at least in languages I can read, reveals a surprising amount of difference in the website taxonomies. Moreover, I suspect that the majority of retailers considering taking their e-commerce international are not likely to be selling furniture.
A reasonable alternative is another Swedish retailer that’s gone global: H&M. A look at the “Ladies” section of their site suggests that it is very consistently structured across all the 61 countries in which they are operating. Moreover, even the non-transactional countries seem to have pretty much the same taxonomy. (There are evidently 2 slightly different releases of the site running globally):
This suggests another methodology for comparing SEM costs, which in contrast to using branded terms, uses generic terms: simply take the second-level taxonomy from H&M’s site around the world and, once again, use the Google Keyword Planning tool to compare costs. This should have the advantage that not only has someone else pre-solved the translation issue for us, but the terms should be relatively free of cultural bias too – a shirt is consistently a shirt everywhere.
(Note that I have additionally separate terms such as “Jackets & Coats” into two separate terms “Jackets”, “Coats”; I wouldn’t like to swear in court to the precision with which I achieved this task in Japanese, although it was fairly simple in most other languages.)
Before showing the graphs, I should note that of course there will still be some bias: coats are likely to be less popular in hot countries for example, and I have a sneaking suspicion that only the British wear socks. By once again weighting the averages according to the popularity of the search terms, I hope to have minimised this bias. There’s also a small secondary advantage from using a Swedish retailer: their original taxonomy will have been designed in Swedish, which removes the skew-to-English generally inherent this approach.
So what does it show?
I suspect the apparent cheapness of Japan, Korea and probably Russia reflect the fact the Google is not the dominant search engine in those countries. However in the case of Korea, especially, there is a suspicion that this may be compounded by H&M making a poor job of translation – the search volumes identified by Google are surprisingly low in a country where they do have at least a 9% share. (The H&M site is not transactional in Korea, so they maybe haven’t put too much effort into optimisation.) As an aside, retailers localising their sites might want to make use of the relative popularity of the translated terms they choose as a sanity check on the translations.
Elsewhere we see a macro-trend: from an SEM efficiency perspective, English language countries are much tougher than European countries. This might be another helpful factor when attempting to justify the costs of translation: not only will conversion rates probably improve, and SEO definitely improve, but it may be cheaper than expected to drive visitors directly via paid methods.
Comparing the approach from the previous article, focussing on international brands such as Nike, Samsung or Lego, with this one focussing on translated menu items such as Jackets, Trousers or Shirts, we see the following:
Pleasingly the results are pretty consistent between the two methods. I’d conjecture that the major inconsistencies can be explained as follows:
Belgium and Italy: e-commerce is surprisingly immature compared to general use of the internet. Searches for branded terms are inherently more likely – and therefore expensive – as online research is more prevalent than online shopping. Note that the generic results for Belgium were based only on Dutch, which probably also affects the data (and possibly contributes to the opposite effect appearing for NL).
Korea, Russia and UAE: possible poor translation by H&M. In Russia there is also the likelihood that internet users are much more likely to be searching for international brands, which may well be the explanation for the UAE also.
The overall conclusions from these two studies (see also ‘Global SEM Costs’) suggest that: