Firefox Retention

Ken Kovash

8

Several months back we discussed a project called Funnelcake. The goal was to begin to understand the path of Firefox users from download all the way through to long-term usage and to see if there were any pain points along the way for our users.

In terms of numbers, we found that 57% of downloaders successfully installed Firefox, and of this smaller cohort, 49% were retained as active daily users thirty days later. Presently, we’re about five months removed from that original time period, so we thought it would be interesting to see what those numbers look like today. In other words, of the group of people who downloaded and installed Firefox on October 4, 2007, how many are still actively using the browser in February 2008?

You’ll see in the chart above that there has been very little drop-off in usage over the past few months. While 49% of installers were still using the browser thirty days later, that percentage has decreased only to 42% nearly five months after the fact. Wow!

This means that Firefox is sticky, really sticky. It means that if someone used Firefox shortly after installing, they’re likely still using it regularly (or daily) five or six months later.

Here’s another way to look at it: five months after users downloaded just two locales (en-US & de) of Firefox on a single day (24-hour period), we see approx. 40,000 of these users each and every day. That’s a mind boggling stat, and only possible because of the amazingly hard work of thousands of contributors.

8 responses

Post a comment

  1. Marcelo Wolfgang wrote on ::

    That’s a impressive mark, but I really like to know what happened to the people that downloaded and don’t even installed it … it’s weird you have to agree with me

  2. Justin Dolske wrote on ::

    It would be interesting to compare these results with another Funnelcake sample… EG, do we always lose ~30% of users over the first 60 days, or were there other factors specific to October and November that contributed to that?

    In particular, I was thinking of FF updates released +14, +28, +43, +47, and +116 from the Funnelcake date. I guess there’s isn’t an obvious correlation, though, now that I look.

    Also interesting that the burp in the +80s matches up with the Christmas holidays.

  3. kkovash wrote on :

    Justin,

    thanks for the comments. I didn’t fully explain it within this post, but we did adjust the daily usage numbers according to our findings here: http://blog.mozilla.org/metrics/2008/01/03/what-factors-affect-firefox-usage/

    So, for the drop-offs in AUS numbers that we typically see with updates being released and with holidays, we did “smooth” our the daily numbers that you see in the graph (e.g., that burp around Christmas was actually much more extreme when looking at just the raw data).

    With that in mind, this funnelcake sample should be extremely representative of other samples. We’re continuing to consider further funnelcake editions and other methodologies to both confirm and enhance our findings.

  4. mawrya wrote on :

    I think I might have a clue about what happens to a large number of the people who download but never installed Firefox.

    If you visit the download page

    http://www.mozilla.com/en-US/products/download.html?product=firefox-2.0.0.12&os=win&lang=en-US

    I know lots of people who would get lost on step 2. Where is this mysterious “continue” button that they should click? I tried downloading in both IE6 and IE7 and the process didn’t seem to match at the instructions exactly. Does the download page sniff the browser and change instructions based on this info? It should. And the instructions should be exact and dead simple. The page is close but could be made even better. Its a shame to think that people get to the point of downloading but just can’t figure out how to do the install due to confusing instructions.

  5. jon wrote on :

    It certainly ‘sniffed’ my Mac. I have installed FF on dozens of computers, and I have never had a problem, but I am reasonably PC-literate. I just find it hard to believe that people would bother to download something, and then not go through what is a simple process of installing. Is there some way the installer could be made even more idiot-proof? I doubt it.

  6. Rob wrote on ::

    Tracking retention is not fool-proof. This is not ie. What I mean is take for example – me, I’ve downloaded firefox numerous times and have uninstalled it numerous times for various reasons. Sometimes I’ve reformatted my computer or installed a new os. This doesn’t mean that I’m not still a firefox users. So it really is impossible to accurately say the % of installers that are still users is XX% when I’ve uninstalled firefox at least 3 times on 2 of my computers. So you now have ‘6 users’ that no longer are using firefox after the original install. The only way to accurately track is to have usernames or have users create an account when downloading, but of course that would prevent at least 20% of people from initially downloading. With that it really is impossible to track retention, again I’ve uninstalled firefox at least 6 times but I still use firefox daily…

  7. Abhisek wrote on :

    Impressive work, but personally I dont think this is fully proved.
    See, how can we mark a user as regular?if we put a curve, then the percentage you guyz are calculating is actually the percentage of downloaded package.
    I personally use 3 pcs, one in home, 2nd one in office, and one of my friend’s.Personally I use FFox, but other guyz may not be using the same.Sometimes my frnd, or my little sis uninstalled FFox.
    As a result I have downloaded again.
    Last month(sept,2008)I have downloaded firefox 7 times in Win and 2 times in Linux platform.
    If u try to put this data in graph it will surely come as increase in retention.But definitely n unfortunately its not the truth.

  8. Robert M wrote on :

    You could avoid duplicates by requesting the user to enter their email address or unique identifier before downloading (without the creation of an account). It wouldn’t be completely accurate, but it might be helpful.

Post Your Comment