Basil’s Bodacious Blog

We’ve seen reports from the field and blogosphere that Firefox 3.0.1 has somehow broken compatibility for many add-ons. This is indeed not the case. When add-on authors publish their add-on they declare the compatibility range for an add-on. It typically looks like “2.0 to 3.0.*” - meaning that at minimum, this add-on requires Firefox 2.0 and works with any maintaince release of Firefox 3.0. (End of the compatibility range is also called the maxVersion).

Some add-on authors have erroneously used “3.0″ as their maxVersion so when users with these add-ons upgrade from Firefox 3.0 to Firefox 3.0.1, the add-on gets disabled. We encourage these add-on authors to use “3.0.*” as their maxVersion.

We have verified that the majority of add-ons hosted on AMO with Firefox 3 support have “3.0.*” in their compatibility range but there are many sites from where add-ons can be acquired.

This minor revision to AMO includes some fixes to some vexing problems as well as a few ease-of-use features. We’ve added a “slider mechanism” to the front page so that you users iterate through five of the featured add-ons.

Some of the bugs resolved include the dreaded “version bumped add-on will not install”, empty review pages (sometimes), and ability to see sandboxed versions of a public add-ons.

For localizers: You can now upload language packs for applications.

For Add-ons Authors: Previously, if you wanted to link to the XPI file hosted on AMO, you had to update your links for every version released. Now you can link to the latest downloadable revision of your add-on by using the following URL: https://addons.mozilla.org/firefox/downloads/latest/{Addon-ID}. This will redirect to the latest available XPI file for your add-on.

See the full fix list.

I’ve previously described the process we use on AMO for rotating the featured add-ons. As we start a new month, we’ve rotated many of the add-on based on their performance. We examine how many additional active add-on users have been added to the Firefox add-ons user base (in absolute numbers) and the relative user retention performance — do users who downloaded the add-on continue to use it? Those with better overall performance stay on the list, the other slots on the list get rotated to uncover new, great add-ons. In this rotation we had 21 add-ons get rotated out and 23 come in. You can see the resulting list for July 2008.

This is a first in a series of blog posts that will try to articulate some background about add-ons in order to explore potential product directions for the Mozilla add-ons site (AMO).

First some facts and figures to help set the context and scale for Firefox add-ons.

• There are more than 180 million active monthly Firefox users (more than 60 million daily)
• There have been more than 800 million add-on downloads since the inception of AMO
• There are about 150 million active daily add-ons (Each add-on downloaded from AMO pings once a day)
• AMO hosts about 5600 add-ons and 800 themes
• We believe that there is a sizable number of add-ons not hosted on AMO (more than 1000?)
• The AMO site has about 4-5 million page views and more than 750,000 user sessions daily
• Every day 10-20 blog/news articles are published reviewing or mentioning some Firefox add-ons

Basically, there is a TON of user and developer activity here.

Why are add-ons important?

• They are core to the Mozilla’s mission to offer “choice & innovation”
• Customization is a key differentiator for Firefox. Huge momentum versus Internet Explorer, Safari (no official site), and Opera
• Add-ons have become a platform for innovation and experimentation and it’s relatively easy to get started
• Many startups have made their bet with the Firefox user base as the entry point
• Major web properties invest in building & distributing add-ons
• We believe that Firefox users who have installed add-ons tend to be more loyal

AMO’s Core Principles

Given the above, the AMO site’s function is to present end users with add-ons, keep them up to date and support the developers who publish these add-ons. The AMO development team focuses its efforts around four main guiding principles:

• Help users find and discover add‑ons in their local language
• Ensure the highest quality for what gets published
• Create a satisfied ecosystem of add-on developers, editors and translators
• Create the most comprehensive add-ons directory on the web

In my next posts, we’ll explore the various constituents (End Users, Developers, Translators & AMO Editors) that make up the add-ons ecosystem and their unique needs.

(Note: we should recognize that even though the AMO site only supports official Mozilla products: Firefox, Thunderbird, SeaMonkey and Sunbird. To look at the entire ecosystem, we need to examine the many XUL-based applications and their support for add-ons. Many applications have their own dedicated add-on site such as the ones for Flock, Songbird, eMusic, and Komodo.)

Where have we been - where are we going?

We’ve seen some major changes to the AMO site over the last six months. We rolled out a major version which we probably should have called AMO 4.0 but didn’t. We then changed to a rapid iteration cycle and have released six minor revisions.

We’ve updated the latest AMO Roadmap to reflect some of the changes we have planned for Q3/Q4 2008.

I’ll be following up shortly with a series of blog posts about add-ons and AMO in general and we’ll revisit the roadmap as part of the feedback and discussion around the future of AMO.

Just a quick reminder that the deadline for the Extend Firefox 3 Contest is coming up quickly on July 4th. This contest includes some great prizes up for grabs and has categories for new, updated and music-oriented add-ons. Read more about it.

Mozilla aspires to be as public as possible with its data and stats. (See Blog of Metrics) We are trying to do our part with AMO and share information about activity we are seeing on the site. Fligtar has two great posts about site metrics during last week’s Firefox 3 launch and the great work that the AMO editor community has done.

We also encourage add-on authors to publish stats about their add-ons. This can be enabled for AMO-hosted add-ons with a flip of a setting accessible at the bottom of your stats chart.

You can view all public add-on stats such as download and active user counts @ https://addons.mozilla.org/statistics/

Happy analysis!

The AMO team has been cranking hard for the past several weeks to bring you the latest batch of goodness. This week’s release includes some very cool end user and developer features, just in time for Firefox 3.

Theme Browser
Ever find it hard to see a bunch of Firefox themes at a glance? We did too and so we created the Theme Browser.

You will now be able to quickly see many theme previews at once, switch between categories and re-order to your heart’s content.

Advanced Search
Just click on the new “Advanced” tab at the bottom of the search panel and you’ll access the advanced search options where you can specify your search using all sorts of parameters.

You can limit your searches for add-ons to selected types, platforms, application version ranges, when it was last updated; then order the results by popularity or other orderings. One capability which may not be so evident is that leaving the search term field blank, will show you all the add-ons that meet the specified criteria, so you can easily find all Firefox 3-compatible themes (as depicted above) or the recently updated Firefox 3-compatible add-ons, etc…Of course, you can bookmark these custom searches to quickly get at your unique search results.

Developer Tools
The last area with new AJAXy flavor and lots of changes are the Developer Tools. A near complete rewrite, it is currently deployed alongside the older version until the full migration is complete. It has been described in detail in a five part series by Fligtar. With this release of AMO, you can try editing your add-on properties, authors, etc… and let us know if you find any problems. (The add-on submit process and the image preview manager are still under development.)

All the Rest
Besides that, the team has fixed a myriad of stuff including:

• Auto-approval of all users reviews. All user reviews are made public immediately. Users and add-on authors can flag questionable reviews for moderation by AMO editors.
• Optimized search performance and moved to a full text search module
• Changes in the search algorithm so that exact name matches show up first
• Ability to deliver add-on updates to language packs
• An updated site privacy policy
• Several new localizations, now at 27 total including: Catalan-Valencian & Hebrew
• The inclusion of experimental add-ons on user info pages

As I’ve previously described, we’ll be rotating the featured add-ons on a periodic basis. This evening we rotated out some category-recommended add-ons from the featured list and replaced them with others from the category-recommended list that haven’t appeared on the front page of AMO yet. Take a look at the latest batch. (You’ll be pleased to hear that more than 95% of the add-ons on this list are Firefox 3 compatible!)

Some of the metrics that I’ve been looking at attempt to measure add-on effectiveness and performance on the AMO site. For example, how much does having an add-on appear on the featured list help? What does it do from a number of add-ons downloaded perspective? What does it do to the active user count for that add-on?

For add-on downloads, I’ve been calculating the change in the 7-day moving average (this tends to eliminate day-of-week fluctuations). I compare the value before and after being on the featured list and note the percentage change. You would assume that with more “exposure”, the 7-day average for featured add-ons would universally increase across the period. But in fact, it doesn’t for all add-ons.

• Some of the newer add-ons to the AMO site (low counts to begin with) saw a tremendous increase in download counts, even upwards of 500%.
• Some add-ons had barely any change in their downloads, these were mostly “older” add-ons with reputations and history where the majority of downloads come from users searching for these add-ons by name. Adding them to the featured list did barely a dent to their download counts.
• Some add-ons’ download counts actually headed south after being on the list. I attribute it to either a lack of interest in the functionality, bad description or because having it featured did not affect the overall downward spiral for the add-on.

As for changes in active user counts, I looked at the total change in active users before and after the featured period. This is a great metric because it is an indicator of how many users have tried it and still have it installed. It doesn’t necessarily mean that the users are using it on a daily basis but at least they didn’t uninstall it. All add-ons on the list had large positive values in absolute terms but what’s keys is the relative change compared to the existing active user base. Some add-ons saw as little as 3% increase in their active users, others saw 354% increase.

Finally, I looked at a metric to measure likeability or retention (active users/download). For those users that downloaded the add-on, how many decided to keep it. That ratio ranged from 9.5% to 48.0%.

I hope this provided a little more metric science insight to the building of the featured list. I’m hoping we can do a similar analysis on the production AMO site to uncover add-ons who are movers-and-shakers and help our broader user base discover cool new add-ons and keep them as loyal Firefox users.

Mark Finkle & Dave Townsend wrote in their blogs (& here too) about the new add-ons newsletter. Receive timely news about changes to the AMO site as well as add-ons related news. You can subscribe here.