Do you setup separate projects per language on a multi-language website?
Will most of your tests be site-wide or will they be langauge specific?
Does your site easily support different look and feel for each region? (is your team built to support different websites per region)
If a site-wide experiment is a winner over-all but flat or slightly negative in one region, will you roll it out globally?
Are you comfortable enough with URL targeting to segment your experiments effecively?
Do you have another analytics tool that you will use to consolidate the experiments from different languages so you can determine an overall impact?
IMO, one project per "code base". This gives the best flexibility and re-usability. But you should consider your team's abilty to support the system in either configuraiton. If your site is identical other than language, use one Optimizely project for all languages and make language available as a segment that you can use when you want to see the results broken down by language.
I do have to test the same page with two variations (50% of audience each), in many different languages.
Language is specified in the url with a proper code such as it, en or similar, just after the http://.
The test concern the effectiveness of static jpgs vs animated gifs, same positions in the page, same language on the pics (hashtags in visuals).
What would you suggest?
I took a look over your questions and I have a few suggestions for you on how best to approach to setting up Optimizely with your web site.
For your original question:
I recommend using one project/snippet on all the different language pages of your site instead of using a different snippet for each language. Here are a couple reasons for this approach:
- Experiments can run site wide on features/sections that that impact all languages, for example navigation bars or footers.
- Using one snippet for all your experiments will also allow you to compare metrics from one language against another language. This can provide insight on whether an experiment resonates more with visitors using one language vs another.
- You can also easily set up your experiments to target just one section of the site, which represents one language.
I do have to test the same page with two variations (50% of audience each)
No. Optimizely allows you to use as little as one variation if you are looking to make a change to your site or several variations of your page, you want to compare against each other.
Optimizely allows you to adjust the traffic allocation, which is the percentage of traffic from your site you want to include in your experiment.
The distribution of the traffic to the different variations can also be adjusted to any percentage between 0 and 100 as long as the total amount of all the variations adds up to 100%, see my screen shot below:
Hopefully this better answers your questions.
Two variations: one for jpgs and one for gifs.
Am I supposed to do anything else?