577778 - Mozilla websites should provide geo-specific content in addition to language-specific content

Status: REOPENED

(www.mozilla.org :: General, defect)

Description

[Rajesh Ranjan]

User-Agent:       Mozilla/5.0 (X11; U; Linux i686; en-US; rv:1.9.0.19) Gecko/2010031706 Red Hat/3.0.19-1.el5_5 Firefox/3.0.19
Build Identifier: 

Currently Mozilla and associated websites show content according to the operating system and language. It should show content related to country specific. For example if from India, I open the page firefox.com it should list first the Indic download instead of anything else! Generally people don't know whether Firefox is available in their local languages. So if page directs to country specific info then it will help local language and native people!

Reproducible: Always

Comment 1

[Rajesh Ranjan]

Anurag jee, I have cc'ed you! This is the bug I was talking about you in the Mozilla Summit 2010.

Comment 2

[Chris More [:cmore]]

Closing old Mozilla.org website bugs due to them not being relevant to the new Python-based Bedrock system. Re-open if this is a critical bug and should be resolved on the new system too.

Comment 3

[Samuel Sidler (:ss)]

This is true, but it might already be done? I know there was discussion before, not sure if it resulted in an actual fix.

Comment 4

[chris hofmann]

I think pascalc was working on geo-ip abilities on the side. if so, maybe this is a dup or could be the tracking bug for that.  we really do need that ability regardless if we are using python or oather based systems.

Comment 5

[Gen Kanai [:gen]]

I should add that we are already doing this (showing country specific content) for Indonesia as referenced in bug 766851, so this does need to be completed.

Comment 6

[Pascal Chevrel:pascalc]

We did have some country-based messages on the old mozilla-europe.org site using maxmind GeoLite database and ther PHP library, the php code we were using was ported to mozilla.com. It should be reimplemented in python by the webdev team if we need it on the current python site, there is a python class they provide http://www.maxmind.com/app/api.

Comment 7

[Chris More [:cmore]]

We'll keep this open for new Product Manager for Websites that will start in August. Language vs country is a good question, but it would more likely require changes to our L10N process. The new Python-based code is still mainly en-US only, but my team and Pascal are working together to get L10N moving along. Many other locales are still getting the old look/feel of the legacy PHP/non-responsive Mozilla.org design.

Comment 8

[Rajesh Ranjan]

Almost four years on, this bug is still waiting, please!

Comment 9

[Chris More [:cmore]]

(In reply to Rajesh Ranjan from comment #8)
> Almost four years on, this bug is still waiting, please!

Can you help us with defining some specific use cases for this? Mock ups of the download page or what text/buttons would be displayed with geo-specific content would be helpful.

View this page in a non-Firefox browser to see what the current download page looks like:

https://www.mozilla.org/en-US/firefox/new/

Thanks!

Comment 10

[Rajesh Ranjan]

Sorry for replying late!

When I filed this bug, I felt a need that when anybody need to download Firefox s/he should be shown the download language option from the region s/he belongs. Means if I belong to Hindi region, I should be shown Hindi Firefox download option and along with the same it should be mentioned that lot of different language from India is also present. Something like this. 

Also, Other languages like section is appearing at the bottom. Language change option should come at the beginning of the page.

Comment 11

[Chris More [:cmore]]

(In reply to Rajesh Ranjan from comment #10)
> Sorry for replying late!
> 
> When I filed this bug, I felt a need that when anybody need to download
> Firefox s/he should be shown the download language option from the region
> s/he belongs. Means if I belong to Hindi region, I should be shown Hindi
> Firefox download option and along with the same it should be mentioned that
> lot of different language from India is also present. Something like this. 
> 
> Also, Other languages like section is appearing at the bottom. Language
> change option should come at the beginning of the page.

Hi Rajesh.

What if I were to come to the Firefox download page with an en-US Internet Explorer browser in India? Should I be automatically offered Hindi?

We are seeing 93% of visitors from India with their non-Firefox browser set to en-US. If they have their browser set to Hindi, we offer them Hindi. If they have en-US set, we offer them en-US.

Firefox is available in these 4 Indic languages: Punjabi, Bengali, Gujarati, Hindi

What if we were to still default to the browser's language preference, but also immediately highlight that Firefox is available in those 4 other languages?

I could pretty easily do an A/B test targeting only visitors from India and then the goal of the A/B test to see what language people click on the most when others are offered. The languages would include English, Punjabi, Bengali, Gujarati, Hindi.

Thoughts?

Comment 12

[chris hofmann]

> Firefox is available in these 4 Indic languages: Punjabi, Bengali, Gujarati, Hindi

Actually I thing it is this set of Indic languages that we have for Firefox desktop that are reasonable to offer in India.

Assamese
Bengali (India)
Gujarati (India)
Hindi (India)
Kannada
Maithili
Malayalam
Marathi
Oriya
Punjabi (India)
Tamil
Telugu

The number of users for each of these languages is currently extremely low to what we might expect when compared to en-US and en-GB users within India.  It seems that just about everyone using the Internet in India up to this point either prefer English, or can't find these other available languages.   

The next generation of Internet users in India will probably prefer these languages so I agree with Chris More that we should do all that we can to promote the availability of these other Indic languages.  If we can detect a user trying to download Firefox from within India let's offer them all of these languages and see if we can have an impact on the numbers.   This set of languages could have further regional breakdown but our geoIP detection might not be that precise and migration of different language speakers around within India might also miss matching languages to speakers.

Comment 13

[Chris More [:cmore]]

I've started the India A/B lang test yesterday. I should have results by the end of the week.

Comment 14

[Chris More [:cmore]]

I just ran a test for 37,000 Indian visitors.

I tested these variations of the en-US Firefox download page:

1) Original (English button only)
2) Punjabi+English
3) English+Punjabi
4) Bengali+English
5) English+Bengali 
6) Gujarati+English
7) English+Gujarati
8) Hindi+English
9) English+Hindi

* ~75% of visitors choose English Firefox
* ~3.6% of visitors picked a non English Firefox
* The order of the languages impact conversion rates

The top clicked non-Firefox downloads:

* Hindi 4.5%
* Punjabi 3.7%
* Bengali 3.5%
* Gularti 3%

So, about 95% of visitors who decided to download Firefox (~78%) during this experiment choose English even when presented other languages. This was similar to my estimate of 93% in comment 11.

Another follow-up test would be to present all languages on one screen to see if we still see the 95% English preference among Indian visitors.

Comment 15

[chris hofmann]

> Another follow-up test would be to present all languages on one screen to
> see if we still see the 95% English preference among Indian visitors.

Yeah, that's the best next test.  If we happened to show Punjabi+English or English+Punjabi to users in Southern India our success rate is likely to be low, or presenting Hindi in Tamil speaking regions likewise.

giving all Indic language choices to all users within India adds some complexity but is the best chance at providing good matches to what people might be looking for or might actually perfer.

The other thing to study are changes over time.  Like mentioned before, the first wave of internet users have had a high preference for English.  If internet use in India continues to grow as expected it will need to reach a wider non-english speaking audience.  When that change happens as been hard to predict.  The thing we need to measure and watch is if there are changes and growth in download and use of these Indic langauge versions over time.  It would not take too much for the number of Hindi users to double over the next months or year.  Once that growth starts it may accelerate quickly.  We need to stay ahead of the curve on this and not be reactive and try and catch up when behind.

Comment 16

[Chris More [:cmore]]

Thanks for the feedback chofmann. I will get the second test set up now.

Comment 17

[chris hofmann]

The one other benefit of providing the list of supported locales in places India might be a bit harder to measure.   If a user sees that extended list they might select and english version, but the might develop a greater affinity for Mozilla knowing that we support and promote languages important to preserving the cultural identity of the region.  This may be confusing, but the way to think about is that people can be supporters of programs like this with out being participants.

http://www.thenational.ae/news/world/south-asia/eager-to-preserve-language-indians-from-kerala-embrace-wikipedia

Developing greater awareness and affinity for Mozilla's support for language would be a win for us, even if we didn't get a single extra download of the Indic versions.

One possibility for testing this theory might be to survey a small sample of people that hit the download page to ask them why they made the language selection they did, or why they chose english and none of the Indic language versions were of interest to them.  Maybe the survey could happen while the stub or installer is downloading.

Comment 18

[Rajesh Ranjan]

Thanks Chirs More! Thanks a lot Chirs Hofmann for supporting and writing in detail here.

http://www.forbes.com/sites/saritharai/2014/09/18/india-wants-to-build-massive-digital-infrastructure-to-cover-800-million-rural-citizens-by-2019/

"India Wants To Build Massive Digital Infrastructure To Cover 800 Million Rural Citizens by 2019"

India is moving fast towards e-Gov and use of internet is spreading in small towns as well. We can hope a good stats.

Comment 19

[Chris More [:cmore]]

I'm re-running the test with these variations:

1) Original

2) Original + 4 alt Indic languages under download button

3) Remove download button and list English and 4 Indic languages links

4) Remove download button and list 4 Indic languages and English links

I'm targeting 100% of Indian visitors to the en-US page that have their browser set to English.

Comment 20

[Chris More [:cmore]]

Test complete!

Here's the conversion rates using the variations described in comment 19. The highest converting variation of the variations above was variation 4. Variation 4 was when I showed "Hindi, Punjabi, Bengali, Gularti, and English language all one one screen with English last. The conversion rate of visitors who downloaded a non-English Firefox even through English was last in the list was 3%. 97% of all Indian visitors to proceeded to download English Firefox even when all languages were presented at the same time and English was last.

This confirms the original assumption that *most* visitors from India even when presented with alternative languages, prefer English. I don't believe it is worth the investment of highlighting and maintaining a per-geo preference list for such a small percentage of visitors. There could be other geos that have a higher alternative language opportunity. It appears the current set up is working 97% of the time for India and highlighting alternative languages on the download page only satisfies 3% or less of visitors. If visitors want another language than their current browser's language, they can still proceed to the "System & Languages" link that shows all languages in the world, i.e. giving users choice.

Comment 21

[chris hofmann]

Hi Chris,

Are the some mock ups so we can see what the experience looked like?

Actually Hindi, Bengali, Telegu, Maraithi, Tamil, Gujarait, Kannada, Punjabi   would be the ordering based on number of speakers that might have the highest reach or need for alternate version.  But at this point I'm not sure swapping in Telegu and Marathi would have made a big difference in the numbers.  

http://en.wikipedia.org/wiki/Languages_of_India provides good "speaker" counts, then the numbers must be adjusted for literacy rates and expected internet use expansion for region and language.

Not investing in this means we are still missing out on the opportunity to present a large number of non-English speakers within India that are expected to come online in the next years.

http://en.wikipedia.org/wiki/List_of_countries_by_English-speaking_population
shows only about 18% of the 1.2 billion people in India speak English.  That still a big number and a big opportunity, but still much smaller than the billion non-English speakers that we should be trying to reach.

I'd suggest we move forward with trying to capture those 3 pct., then with some community word of mouth see if that number grows; and with some social media and other marketing see if we can be tagged as "the" browser for Indic non-English speaking users, or at least generate a conversation about which browser offers the best experience for those billion non-english speakers in India that are coming on-line.

But then again I've always been interested in crazy things like growing firefox users when starting out with just 3% and trying to grow it to something much bigger, and getting the conversation focused on the match ups and who provides the best browsing experience.

http://www.forbes.com/2004/10/04/cx_pp_1004mondaymatchup.html

http://www.techweb.com/news/163702025/mozilla-readies-alpha-release-of-firefox-upgrade.html

What other market is there such a big opportunity waiting to be tapped as users come on-line where the user language choice is not so obvious or clear?  This really is good fruit for a 2015 goal and a place where we should be investing to try and move the needle.

Comment 22

[Chris More [:cmore]]

What would be helpful is if we could start a spreadsheet where we list out all the countries and the order of the primary languages spoken in those countries. It would help get a scope on the overall challenge of this project.

Let me get some screenshots of the test.

Comment 23

[Jobava]

Chris, do you still have the data from those tests you ran last year?

Comment 24

[Chris More [:cmore]]

(In reply to Jobava from comment #23)
> Chris, do you still have the data from those tests you ran last year?

Yeah, I can find out something things from the tests last year. I included some of it in comment 20. Do you have specific questions?