March, 2009

process / results — 11 Comments
31
Mar 09

Why People Don’t Install Firefox – Part I

Installing a piece of software can be a challenging process for the typical person to traverse.  For example, with Firefox, we know that each day there are more than 50,000 people who fall into the following group – they complete the download process (receive the full data/file), but don’t then complete their installation of Firefox.

We feel we have a moral imperative to figure out what’s going on here.  Doing so provides us an opportunity to both add to the general knowledge base of the software industry and make significant strides in the Firefox user experience (and to boot, improving Firefox adoption rates isn’t a bad auxiliary outcome).

So, let’s start by outlining the exact steps that a software user needs to traverse after a successful download:

  1. Does a user know they need to install?
  2. Does a user remember where he/she saved the file?
  3. Does the user open/run the file and avoid hitting “cancel” during extraction?
  4. Does the user make it through all steps (either 3 or 6 steps in the case of Firefox) of the installer without hitting “cancel”?

Thanks to an important initiative launched last week, we’re now able to put some framework and answers around these questions.  On Thursday of last week, we shipped a modified version of Firefox with a different installer (en-US*Windows only for a single day), allowing users to leave feedback when canceling out of the installer.

step1_cancel_red

Based on the results (coming in a follow-up post), we know that of our original cohort of 50,000 people, approximately 10,000 users fall under category #4 outlined above (i.e., they click on that cancel button).  That means steps #1-3 describes the experience/actions of about 40,000 people.

Thinking in terms of big picture implications, there are some possible installer changes that we can consider to remedy the issues encountered via the installer itself.  Perhaps more paramount – what can we do to help users from steps #1-3?  Content and design of the Firefox download page?  Any other ideas?

To be sure, any action that we take here could be wildly impactful.  Improving the experience for just 2,000 or 3,000 people from our original cohort would translate into upwards of 2,000,000 additional Firefox users annually.

Uncategorized — 10 Comments
19
Mar 09

What is Firefox’s Market Share?

In recent weeks, we’ve seen either major updates or new launches from a couple significant sources of browser market share.  Thus, we thought this would be a good opportunity to provide a high-level overview of a few sources that are out there, what their samples look like, and the ability of each to slice and dice data.

Below you’ll find four different sources of browser market share data.  Please note this isn’t meant to be a comprehensive list (if you know of other good sources, please leave a comment).  I’ve sorted the following by sample size, largest to smallest, based purely on my own best guess.

1) Gemius

Gemius describes themselves as the “largest online research agency in Central and Eastern Europe”.  Even though their data is concentrated on just one region, their sample size is stellar.  Looking at any report, you can quickly see the exact number of page views that their data is based on, e.g., browser market share for last week in Poland is based on nearly 10 Billion page views.  Looking across different countries, it appears that their monthly sample size is greater than 50 Billion page views.  Better yet, all of Gemius’ market share data is free and publicly available.

gemius_example

2) Net Applications

Net Applications describes themselves as “a leading source of applications for webmasters and eMarketers”.  Their browser market share data has been published for many years and has been often relied upon by the business press.  Derived from site visitors to their network of live stats customers, Net Applications’ data is based on 160 million visitors per month.  Most of their data is available publicly, and for a small subscription fee, you can also drill into their data by region/country, by week, and in several other ways.

net_app_example

3) StatCounter

StatCounter appears a bit similar to Net Applications in that they’re a web analytics firm publishing market share reports based on data from across their customers’ web sites.  While StatCounter has been around for many years, they just recently launched their browser market share report.  They present a super slick (i.e., cool) interface where you can instantly adjust the date range, select a region/country, and choose a bar or trended line chart.  And it’s all publicly available.  They describe their sample as follows, “Stats are based on aggregate data collected by StatCounter on a sample exceeding 4 billion pageviews per month collected from across the StatCounter network of more than 3 million websites.”  So far, I’ve noticed a few inconsistencies when drilling down by region or country (and the data only dates to last summer), but otherwise, it’s good to see StatCounter recently release this service.

statcounter_example

4) AT Internet Institute (formerly Xiti)

AT Internet Institute describes itself as a company that “enables an integral analysis of websites, intranet and mobile sites”.  They have been publishing browser market share reports for nearly five years, and they collect their data in a way similar to the providers highlighted above.  For their most recent report (released last week), AT Internet Institute describes their sample as a “cross-section of 122,099 websites”.  I believe the majority of these sites are based in France, and the vast majority of the traffic comes from European internet users.  While you can’t drill down into their raw data, their quarterly reports do a fairly good job of analyzing and interpreting interesting trends.

xiti_example

Uncategorized — 3 Comments
16
Mar 09

The Global Growth of Firefox

This post may only be exciting for those who enjoy spending their free time coding in a statistical program (e.g., R, Stata), but I wanted to share an output courtesy of Blake Cutler.  Taking per country browser market share data from Net Applications since the beginning of 2008, Blake simply gave a “sum” command (i.e., summarize) for both January 2008 and February 2009:

market_share_sum1

You’ll notice a few interesting trends:

  • The number of countries in the sample has increased over time.
  • The mean (or average) Firefox market share, on a per country basis, showed a very healthy increase in the past year.  The means of 20.1% and 27.7% are greater than what you see here because they are not weighted in any way (e.g., by region or internet population size).
  • Correspondingly, the standard deviation has increased as well.
  • The “max” increased from 47.2% to 63.1% (I believe that’s Indonesia).
plans / process — 3 Comments
4
Mar 09

More Improvements Coming to Firefox Web Sites

Last fall we initiated the first ever multivariate optimization testing at www.mozilla.com.  At the time, we drew the following conclusion from one of our tests:

How does this change our business/creative decision making?

As some of us already guessed, it appears that it’s the download button itself that matters most here.  Thinking about the size, shape, color, and placement (among other characteristics), our findings indicate that future testing could reveal surprising – and positive findings – based on changes to the download button.  In other words, it wouldn’t be a stretch to imagine that future tests, that adjust such characteristics of the download button, could easily translate to a 2% lift (or much greater), and in turn, a substantially improved experience for millions of new Firefox users each year.

We have plans in place to eventually roll-out those types of tests with regular frequency later this year (stay tuned), and in the short term, we’re planning a few strategic optimization efforts at a few critical Firefox related touch points – support.mozilla.com, Get Firefox in your language, and the main Firefox product page.

David Rolnitzky just highlighted our plans regarding that last touch point.  Among the changes being tested:

  • If an existing Firefox user is trying to figure out how to upgrade to the latest version of Firefox, the process should be clearer.
  • If a user is on IE, he/she will see content more relevant to their thoughts and potential interest in downloading Firefox (below).

ie_version_of_fx_product_page

This is only a test, so while these new page versions seem as though they’ll have a positive impact on users, we’re not necessarily rooting for them to exhibit greater success than our existing page (the “winner” is determined by the actions/preferences of site visitors).  Stay tuned for our findings in a future post.