Spreadfirefox affiliates
Go to the Spread Firefox wiki home
Go to Spread Firefox Affiliates program wiki home
Contents
The Affiliates program
What is it?
The Affiliates program is a Spread Firefox marketing project. Community members, or "affiliates", place "buttons" on websites all over the Internet; on blogs, forums, project websites, you name it.
Buttons come in many different shapes and forms...
Our most popular buttons are the ones that lead to Firefox downloads...
Part of the program involves tracking metrics related to the usage of these buttons. For example, how many Firefox downloads did Affiliate X create this week by having a button on their blog?
Types of metrics currently tracked:
- Hits - Someone clicked a button
- Downloads - A download button led to someone downloading Firefox.
- Points - Download buttons have a point value associated with them. This allows users to collect points.
- Last point change - The amount of points a user gained during the tally. There is no front-end UI for this. It helps to aggregate other stats/lists.
Lists:
- Top 250 - Ordered by the number of downloads over the past 7 days
- Top 5 - Same as Top 250
- Top movers - The top 5 affiliates ordered by (total this week - total last week), where this week is the past 7 days, and last week was between 7-14 days ago.
Points can lead to rewards:
- Featured Foxes
- This program is currently offline.
- Every quarter, 5 affiliates are chosen at random from the Top 250 list.
- Selected affiliates are sent a t-shirt, and their name and point total are displayed on the Affiliates project homepage.
- Upgrade the Web (upcoming project for 3.5 release)
- A special set of download buttons.
- Users who achieve at least 100 points over two weeks are entered into a raffle.
How does it work?
A button has a few key elements
- Button image, <img src>
- Images are hosted by IT on the static cluster
- Destination, <a href>
- Where does the button link to when someone clicks it?
- This URL has some key elements also. This is how the metrics begin.
- Affiliates user ID and button ID are appended to the destination as query params
- ex: mozilla.com/firefox/?from=sfx&uid=555&t=209
- ?from=sfx - self-explanatory
- &uid=555 - this is the account ID of the spreadfirefox.com user that owns this button
- &t=209 - this identifies which button this is
A script parses the access logs of destination websites, looking for affiliates referrals, using the key elements described above
- mozilla.com
- We look here for visitors coming '?from=sfx'
- Each visitor's affiliate ID, button ID, IP address, and timestamp are extracted
- download.mozilla.org
- We look here for downloads that relate to affiliate traffic, using the IP addresses we extracted from mozilla.com logs.
- If a download occurs from one of these extracted IP addresses, the referrals from this IP address are marked as having a download.
- Only one download per IP, per day, can be counted.
All these affiliates referrals are inserted into the SFx database for tallying
- A script goes over the referrals and tallies hits, downloads, and points for each affiliates.
- If a user doesn't have an existing affiliate record, one is created.
- Next, lists like Top 250 and Top Movers are gathered.
Implementation details
- log parse script
- Perl script
- runs on dm-stats01
- run.pl
- written by Alex B
- outputs SQL INSERT queries for each affiliate referral record
- a cron job picks up this output file and runs the queries on the SFx DB
- Tally scripts
- PHP code
- runs within Drupal framework
- sfx_affiliates.module
- most recently rewritten by Alex B
Ideas
- Have all buttons go through a lightweight jump host, and be redirected to their destination. This would make the log parsing easier and more efficient.
- Track metrics on sfx-images.mozilla.com, this would tell use how many views buttons get