Example Ranking Methods for Your Feeds

This post will show you clear examples for all 4 of those use cases, but first let’s go over the basics of ranked feeds.

Ranked feeds allow you to define a scoring formula for your activities. Within the scope of the ranking method you have access to any of the custom attributes you’ve added to your activities. Here’s a very basic ranking method:

The above uses a simple linear time decay on the “time” field of the activity. It combines the decay score with the square root of a custom attribute called “popularity” sent as part of the activity. If the popularity attribute is missing from the activity payload, we’ll use the default of 1.

Ranking methods are a very powerful tool. The following five examples will give you a bit of inspiration for your own application.

Variations of the following “viral” formula can be used to generate a ranking that is a function of activity popularity and creation time:

Another common use is applications which want to show specific activities when they are about to start. Some use cases include soccer matches, or tickets going on sale, for example.

Here’s an example that shows activities that are about to start higher in the feed:

Sometimes you’ll want to boost promoted or editorial activities. Here are two different approaches:

Formula #1 :

This will ensure that promoted activities will have a slower decay, where promoted is always a float larger than one for promoted activities. We recommend starting with promoted=2 and then tweaking it if that value doesn’t work. Remember: the bigger the promoted value, the slower the decay for boosted activities.

Formula #2 :

This is another alternative that is more aggressive but easier to interpret; for instance, setting promoted equal to 2 makes boosted activities stay at the top at least as long as scale+offset. In general, if promoted is larger than 2, the activity will stay on top of other non-boosted activities for at least scale+offset.

The ranking method shown below uses the “click” and “like” events from analytics on a feed:

Using analytics data makes it easy to rank activities based on engagement metrics such as clicks, hover events, video watches, and so on.

One question we often get is about the difference between ranked and personalized feeds. There are 2 main differences:

Stream’s personalization can be used for:

Personalized feeds are custom built per app and work well for a wide variety of use cases.

Related posts:

This morning did not begin well. I admit to being a bit grumpy due to an unseasonable chill in the air and the fact that my wife of many decades chose to sleep in, leaving me to prepare my own…

I believe that in theory it is possible to “have it all”, but this almost never happens in practice. There are far too many aspects that can result in compromises and sacrifices being made at some…

SpaceX CEO Elon Musk gave a detailed commentary on the future flights of mankind to Mars and shared his vision of what exactly it will take to build the first full-fledged city on the red planet. The…